SLinstall
Jump to navigation
Jump to search
Notes
- these instructions are periodically updated to include items needed for older/newer versions of Linux. They are marked like this: (SL4.2+) means Scientific Linux 4.2 and newer; (SL4 is equivalent to FC3). (FC5 only) means Fedora Core 5; etc.
- obsolete items are marked by the "#" sign at the beginning of the line and sometimes have a comment about the reason for removal.
- typically, we do not "upgrade" machines using the Red Hat "upgrade" function. Instead, we save critical files from the old installation and do a "fresh install" from scratch
- starting with RHEL7, the recommended OS is CentOS7 (instead of SL7).
Disk configurations
The year is 2019 and SSDs are used exclusively, except for bulk data storage, where one used 6-8-10-12 TB HDDs
For reliability, home directories and data disks must use redundant storage - mdadm raid1 or ZFS raid1/raid6.
For non-critical machines, a single SSD seems to be reliable enough to use as a boot and OS disk. But since any storage device can fail at any time without warning, home directories and data disks should use redundant storage.
Note: for data disks bigger than 4-6TB, mdadm raid1/raid6 is no longer recommended because raid rebuild, verification and repair time has become unreasonably long. Instead, use ZFS raid1/raid6 which implements online verification, repair and disk replacement without requiring machine shutdown or OS down time.
- single SSD - 120GB min - single partition for "/", no swap partition (create a swap file if swap is needed) - for non-critical machine with no local data storage (OS only)
- dual SSD - 2x240GB min - all partitions mirrored (RAID1), 30GB "/", rest for /home1 - for daq station with local user home directories and no bulk data storage
- single SSD + 2x6-8-10-12TB HDD - SSD partition: all "/", HDD partition as ZFS raid1 (mirrored) - for daq station with small local bulk data storage
- single SSD + 6-8x6-8-10-12TB HDD - for small storage server machines - for daq station with local home directories and large bulk data storage.
For VME processors:
- network boot - VME-CPU#Network_boot - only option for V7648/V7750, do not use for V7805 (no netboot from GigE), optional for V7865/XVB-602
- USB boot - 8GB USB for V7805, 16GB USB for V7865/XVB-602
Preparation
- save /etc, /var, /root, /opt, (if needed: /usr/local, /tftpboot) by rsync to some data disk (/ladd/data0/root)
- check that "/" partition (it will be overwritten) is different from /home1 and /data partitions
- note the MAC addresses of all network interfaces, add them to ladd00 dhcpd.conf to enable PXE boot into the SL "network installer"
- shutdown
Running installer (CentOS7)
CentOS7 can be installed from vanilla CentOS7 installation media or from a custom USB key build per there instructions: https://daqshare.triumf.ca/~olchansk/linux/CentOS7/
The custom installer makes it easy to use a custom kickstart file (ks.cfg).
Instructions for using the usb-installer:
- disconnect machine from network
- plug the usb-installer into a usb3 port (blue colour)
- reboot machine, select booting from usb (press F8 on ASUS motherboards)
- usb-installer boot menu offers to install CentOS7, go there
- CentOS7 should boot (many messages scroll on screen)
- into graphical mode
- into installer main menu
- all installer options should "happy" except for the "installation destination"
- go to the "installation destination" menu
- unselect all disks except for the SSD where the OS will be installed
- (MOST IMPORTANT: unselect the USB installer disk!)
- select "I will configure..."
- say "done"
- the "manual partitionning" menu will open
- use the "-" button to delete all existing partitions
- select "standard partition"
- click on the "+" button
- in the "Add new partition" dialog, set mount point "/", capacity blank, click "add mount point"
- check capacity (should be full size of SSD), check filesystem type (should be XFS)
- say "done", there will be a warning about absent swap partition, say "done" again.
- in the big useless dialog, say "accept changes"
- should be back to the "installation summary" screen, "installation destination" should be happy now
- after everything is happy, say "begin installation"
- as the installation proceeds, set the password for the root user
- after installation is complete, reboot the machine
- unplug the usb-installer, CentOS7 should boot from SSD into the login screen
- click on "not listed?", login as root
- setup network connection:
- open a terminal
- start "nm-connection-editor"
- click on "+" to create a new connection profile
- select "wired ethernet"
- select "add profile..."
- in "Identity", set "name" to "static"
- in "Identity", check that "Connect automatically" and "Make available..." is enabled
- in "IPv4", set "Addresses" to "manual" instead of "dhcp"
- enter IP address, netmask 255.255.224.0, gateway 142.90.100.18, dns 142.90.100.19, search triumf.ca
- say "Add", then close/quit the network settings
- connect network cable
- network should be up, ping ladd00 should work
- run: yum update -y
- check new kernel is installed: ls -l /boot
- logout and restart (good luck finding these buttons in the gui!)
- confirm correct linux kernel is selected during boot (-229.20, not the original installer kernel)
- login as root, confirm network is up, proceed with the rest of these instructions
Configure SSH
(+CentOS7)
- Login from the console
- restore the SSH keys from backup (/etc/ssh/*key*)
- service sshd restart
- ssh into the new machine as root
- ssh root@localhost, ctrl-C
- ### this is done later from Konstantin's git repository - scp root@ladd00:/root/authorized_keys ~root/.ssh/
- (not needed for SL5.5 kickstart) check that /etc/ssh/ssh_config contains "ForwardX11 yes" and "ForwardX11Trusted yes":
echo " ForwardX11 yes" >> /etc/ssh/ssh_config echo " ForwardX11Trusted yes" >> /etc/ssh/ssh_config
Post installation CentOS7
Set hostname: (use full name, i.e. daq11.triumf.ca)
emacs -nw /etc/hostname
echo "olchansk@triumf.ca amaudruz@triumf.ca lindner@triumf.ca" >> ~root/.forward chmod a+r /var/log/messages chmod a+r /var/log/yum.log
Activate rc.local:
chmod a+x /etc/rc.local chmod a+x /etc/rc.d/rc.local # TL edit systemctl start rc-local systemctl status rc-local
Disable "persistent network names" (DO NOT DO THIS)
/bin/touch /etc/udev/rules.d/75-persistent-net-generator.rules /bin/rm /etc/udev/rules.d/70-persistent-net.rules #shutdown -r now
Configure NIS client (CentOS7)
yum -y install ypbind authconfig echo "NISTIMEOUT=5" >> /etc/sysconfig/network echo "NETWORKWAIT=yes" >> /etc/sysconfig/network authconfig --enablenis --enablepreferdns --nisdomain LADD-NIS --nisserver ladd00.triumf.ca --update ypwhich ypcat -k passwd systemctl restart autofs
- On the master NIS node (ladd00), add this new node to /etc/netgroup, and update NIS maps (cd /var/yp; make)
- Use "system-config-users" to add local user accounts
- enable selinux ssh key login to nfs mounted home directories:
setsebool -P use_nfs_home_dirs 1
Configure NIS secondary server (CentOS7)
Enable local NIS server, make local machine use it:
yum -y install ypserv /usr/lib64/yp/ypinit -s ladd00 ### (/usr/lib/yp/ypinit on 32-bit machines) systemctl enable rpcbind ypserv ypxfrd yppasswdd systemctl start rpcbind ypserv ypxfrd yppasswdd emacs -nw /etc/yp.conf # change "domain XXX server YYY.triumf.ca" to read "domain XXX server localhost" systemctl restart ypbind ypwhich # should say "localhost" ypcat -k auto.master # should work
Punch hole in the firewall: (or "make" on NIS master will complain)
echo YPSERV_ARGS=\"-p 800\" >> /etc/sysconfig/network systemctl restart ypserv firewall-cmd --get-services firewall-cmd --add-service rpc-bind --permanent firewall-cmd --add-port=800/tcp --add-port=800/udp --permanent firewall-cmd --reload firewall-cmd --list-all
- on the NIS master:
- add the new machine to /var/yp/ypservers, run "make -C /var/yp" and also "cd /var/yp; yppush -h newmachine ypservers"
- if using /var/yp/securenets, copy it from NIS master to new NIS secondary server
Enable hourly NIS update cron job (DO THIS AFTER git pull scripts, see below)
cd ~/git/scripts git pull cd etc cd ~/git/scripts/etc; ln -s $PWD/ypxfr-cron-hourly /etc/cron.hourly
Configure AUTOFS (CentOS7)
yum -y install autofs systemctl enable autofs systemctl start autofs ls -l /daq/daqshare
Label Selinux labels
When upgrading non-selinux machines (el6) to el7 (selinux enforcing) the existing user home directories will not have the correct selinux labels and many things will not work, including ssh logins (sshd cannot access ~user/.ssh files).
semanage fcontext -a -e /home /home1 ### selinux has special rules for /home, assign them to /home1 restorecon -R -v /home1 ### apply the new rules to files in /home1 ls -Zd /home1/alpha/.ssh # should say: drwx------. alpha users system_u:object_r:ssh_home_t:s0 /home1/alpha/.ssh
Configure time (CentOS7)
Time server ntpd was replaced by chronyd.
yum -y install chrony echo server time1 iburst >> /etc/chrony.conf echo server time2 iburst >> /etc/chrony.conf echo server time3 iburst >> /etc/chrony.conf systemctl enable chronyd systemctl restart chronyd chronyc sources chronyc tracking
- if desired, edit /etc/chrony.conf, remove non-triumf time servers
Enable automatic system updates (CentOS7)
Disable yum-cron:
rpm --erase yum-cron /bin/rm -v /var/lock/subsys/yum-cron /bin/rm -v /etc/cron.daily/0yum-daily.cron /bin/rm -v /etc/cron.hourly/0yum-hourly.cron
Enable yum-autoupdate:
yum install -y epel-release yum install -y yum-changelog yum-protectbase yum-tsflags yum-versionlock rpm -vh --install http://linuxsoft.cern.ch/cern/centos/7.2/cern/x86_64/Packages/yum-kernel-module-1-5.el7.cern.noarch.rpm rpm -vh --install http://linuxsoft.cern.ch/cern/centos/7.2/cern/x86_64/Packages/yum-autoupdate-4.4.2-1.el7.cern.noarch.rpm #rpm -vh --install https://daqshare.triumf.ca/~olchansk/linux/yum-autoupdate-4.4.2-1.el7.cern.noarch.rpm https://daqshare.triumf.ca/~olchansk/linux/yum-kernel-module-1-5.el7.cern.noarch.rpm systemctl enable yum-autoupdate systemctl start yum-autoupdate systemctl status yum-autoupdate
Disable automatic system updates (CentOS7)
yum -y erase yum-autoupdate /bin/rm -f /etc/sysconfig/yum-autoupdate.rpmsave /bin/rm -f /var/lock/subsys/yum-autoupdate
Configure system services (CentOS7)
- systemctl list-unit-files | grep enabled | sort ### (to see enabled services)
- disable unwanted services:
systemctl disable bluetooth systemctl disable dm-event systemctl disable dmraid-activation systemctl disable iscsid systemctl disable iscsi systemctl disable iscsiuio systemctl disable libvirtd systemctl disable lvm2-lmetad systemctl disable lvm2-monitor systemctl disable ModemManager systemctl disable multipathd systemctl disable netcf-transaction systemctl disable lvm2-lvmetad.socket systemctl disable lvm2-lvmpolld.socket systemctl disable iscsid.socket systemctl disable iscsiuio.socket #systemctl disable
Erase unwanted packages
yum erase PackageKit # bugs users about security updates
Erase unwanted packages (CentOS7)
- PackageKit # bugs users about security updates, hogs yum lock
- perl-homedir # creates unwanted $HOME/perl5
- ModemManager # thinks that all USB-attached devices are modems
- pcp # sends error email to itself, does not work
- abrt # sends email to root about useless crashes, i.e. crash of X when machine is rebooted
- rear # some kind of backup and recovery tool, not clear what it does, but it sends email complaining how it is broken
- bash-completion # "echo $HOME/<TAB>" becomes "echo \$HOME" (notice "\" added before "$") preventing tab-completion from doing anything useful.
yum -y erase PackageKit perl-homedir ModemManager pcp abrt abrt-libs abrt-gui-libs rear bash-completion
Configure external package repositories
yum install elrepo-release epel-release
Configure external package repositories (CentOS7)
EPEL: (addtional packages)
yum install epel-release
ELREPO: (kernel drivers)
rpm --import https://www.elrepo.org/RPM-GPG-KEY-elrepo.org rpm -Uvh https://www.elrepo.org/elrepo-release-7.0-2.el7.elrepo.noarch.rpm yum -y install yum-plugin-fastestmirror
Install packages needed to continue with installation
(+CentOS7)
(these packages are sometimes missing, they are needed to follow following instructions instructions)
(SL6.5: libotf is a dependancy of emacs - SL6.5 installer fails to install it)
yum install ed patch wget git libotf gdisk emacs
Configure TRIUMF packages
(only for machines on the TRIUMF network)
(TRIUMF kickstart usually installs this automatically)
rpm -vh --install http://mirror.triumf.ca/triumf/6/x86_64/Packages/triumf-release-1.4-1.noarch.rpm yum install triumf-ssh triumf-syslog
Configure TRIUMF packages (CentOS7)
(only for machines on the TRIUMF network)
# TL Was rpm -vh --install http://mirror.triumf.ca/triumf/6/x86_64/RPMS/triumf-release-1.4-1.noarch.rpm rpm -vh --install http://mirror.triumf.ca/triumf/6/x86_64/Packages/triumf-release-1.4-1.noarch.rpm yum install triumf-ssh triumf-syslog
Configure Konstantin's scripts
(+Centos7)
mkdir ~root/git cd ~root/git git clone http://ladd00.triumf.ca/~olchansk/git/scripts.git cd scripts git pull
Enable yum version lock
DO THIS ONLY IF NEEDED
yum install yum-plugin-versionlock yum versionlock packagename # yum versionlock rpcbind yum versionlock list # list locked packages yum versionlock delete packagename # unlock given package yum versionlock clear # delete all locks
Configure TRIUMF mirror of yum repositories (SL6)
(only for machines on TRIUMF network)
- if /daq/mirror is available: /bin/cp ~/git/scripts/etc/daq-mirror-SL6.repo /etc/yum.repos.d/
- if /triumfcs/mirror is available: /bin/cp ~/git/scripts/etc/triumfcs-mirror-SL6.repo /etc/yum.repos.d/
- otherwise: /bin/cp ~/git/scripts/etc/triumf-SL6.repo /etc/yum.repos.d/
then disable external repositories:
yum clean all yum-config-manager --disable epel yum-config-manager --disable elrepo yum-config-manager --disable sl yum-config-manager --disable sl-security yum-config-manager --disable sl6x yum-config-manager --disable sl6x-security yum clean all
Configure trusted ssh keys
(+CentOS7)
ssh localhost interrupt by Ctrl-C /bin/cp ~/git/scripts/etc/authorized_keys ~/.ssh/
Configure hardware sensors
- yum install lm_sensors kmod-k10temp kmod-coretemp
- sensors-detect (accept default answer to all questions - press ENTER)
- service lm_sensors restart (to reload the kernel modules)
- sensors (to see available sensors)
If no sensors are detected by standard drivers, follow motherboard-specific instructions at the bottom of this page.
Configure coretemp CPU sensors
On some machines, the coretemp driver for Intel CPU temperature sensors is not loaded after the above steps.
- sensors | grep coretemp ### number of sensors reported should be the same as the number of CPU cores
- if output is blank, add this to /etc/rc.local
emacs -nw /etc/rc.local modprobe coretemp
Configure IPMI sensors
Some machines support the IPMI interface for monitoring the hardware: fan speeds, temperatures, voltages.
- find out if IPMI is supported. Try this:
dmidecode | grep -i ipmi
if output is not blank, IPMI is maybe supported.
- install and enable IPMI software:
yum install "OpenIPMI*" ipmitool service ipmi start ipmitool sensor ### to confirm IPMI is present. If output is blank, do not go further. chkconfig ipmi on chkconfig ipmievd on service ipmi restart service ipmievd restart tail -100 /var/log/messages ### look at messages logged by ipmievd
- (CentOS7) install and enable IPMI software:
yum install "OpenIPMI*" ipmitool systemctl start ipmi ipmitool sensor ### to confirm IPMI is present. If output is blank, do not go further. systemctl list-unit-files | grep -i ipmi systemctl enable ipmi systemctl restart ipmi systemctl status ipmi systemctl enable ipmievd systemctl restart ipmievd systemctl status ipmievd tail -100 /var/log/messages ### look at messages logged by ipmievd
- if ipmievd complains about SEL buffer overflow, clear it manually:
ipmitool sel list ### show ipmi messages in raw format ipmitool sel elist ### show ipmi messages in useful format ipmitool sel elist > file ### save ipmi messages into a file ipmitool sel clear ### clear all accumulated ipmi messages
- useful ipmi commands:
- ipmitool sensor -- read hardware sensors
- ipmitool sel elist -- report all accumulated messages
Configure SMARTD (CentOS7)
Default el7 smartd config files send deficient email notices about disk failures. Overwrite.
/bin/cp ~/git/scripts/etc/smartd.conf /etc/smartmontools/ /bin/cp ~/git/scripts/etc/smartd_warning.sh /etc/smartmontools/ systemctl restart smartd systemctl status smartd
Enable User Disk Quotas (OPTIONAL)
(+CentOS7)
- read http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/6/html/Storage_Administration_Guide/ch-disk-quotas.html
- emacs -nw /etc/fstab, add "grpquota,usrquota" to filesystem options, e.g.:
[root@isdaq00 home1]# grep quota /etc/fstab UUID=5a2aefbd-45db-475e-841e-12ec89220fbd /home1 ext4 defaults,grpquota,usrquota 1 2
- cd /; umount /home1; mount /home1
- quotacheck -cug /home1
- quotacheck -avug
- quotaon -av
- quota system is now active
- increase the soft quota time limit from default 7days to 30 or 60 days: edquota -t
- set quotas for all users (see below)
- setup warnquota:
- create warnquota config file: emacs -nw /etc/warnquota.conf
# values can be quoted: MAIL_CMD = "/usr/sbin/sendmail -t" FROM = root SUBJECT = User %i@%h exceeded allocated disk quota CC_TO = "root" # If you set this variable CC will be used only when user has less than # specified grace time left (examples of possible times: 5 seconds, 1 minute, # 12 hours, 5 days) # CC_BEFORE = 2 days SUPPORT = "root" # Text in the beginning of the mail (if not specified, default text is used) # This way text can be split to more lines # Line breaks are done by '|' character # The expressions %i, %h, %d, and %% are substituted for user/group name, # host name, domain name, and '%' respectively. For backward compatibility # %s behaves as %i but is deprecated. MESSAGE = User "%i" on "%h" has exceeded the allocated disk quota.||Please delete any unnecessary files on following filesystems or|contact the system administrato r to increase your quota allocation:| SIGNATURE = --|automated email from warnquota
- note that %i@%h in the SUBJECT line do not seem to work
- create cron job: emacs -nw /etc/cron.daily/warnquota
#!/bin/sh warnquota #end
- chmod a+x /etc/cron.daily/warnquota
- touch /etc/crontab
Useful commands for managing quotas:
- repquota -a | sort -n -k3 ### show quota of all users sorted by disk usage
- edquota -u username ### open "vi" editor to change user quotas
- repquote -a | grep username ### report quota for given user
- setquota -u username 0 0 0 0 /home1 ### disable quotas for given user
- setquota -u username 50000000 100000000 0 0 /home1 ### set quotas for 50GB soft and 100GB hard
- edquota -t ### change user quota time limits
- edquote -tg ### change group quota time limits
Enable NFS V4 server (CentOS7)
- create /etc/exports. example: (fsid numbers should be unique and increase 1,2,3,...)
/home1 @home_export(rw,no_root_squash,async,fsid=1) /data1 @data_export(rw,no_root_squash,async,fsid=2)
- check the netgroup file
- if using NIS: check NIS netgroup: ypcat -k netgroup
- if no NIS, create /etc/netgroup: @daqmachines (deap00,,) (deap01,,) (deap02,,)
- if no NIS, edit /etc/nsswitch.conf, make the netgrooup line read: "netgroup: files"
- enable things, start them:
firewall-cmd --get-services firewall-cmd --permanent --add-service=nfs firewall-cmd --reload firewall-cmd --list-all systemctl enable nfs-server systemctl start nfs-server systemctl status nfs
Enable NFS V3 server (CentOS7)
ps -efw | grep rpc.mountd # should be running! firewall-cmd --get-services firewall-cmd --permanent --add-service=mountd firewall-cmd --permanent --add-service=rpc-bind firewall-cmd --reload firewall-cmd --list-all
Enable NFS V3 server
- edit /etc/hosts.allow, add or uncomment "mountd: 142.90.0.0/255.255.0.0"
- create /etc/exports. example:
/home1 @home_export(rw,no_root_squash,async) /data1 @data_export(rw,no_root_squash,async)
- check the netgroup file
- if using NIS: check NIS netgroup: ypcat -k netgroup
- if no NIS, create /etc/netgroup: @daqmachines (deap00,,) (deap01,,) (deap02,,)
- if no NIS, edit /etc/nsswitch.conf, make the netgrooup line read: "netgroup: files"
- chkconfig nfs on
- chkconfig nfslock on
- service nfs restart
Then on ladd00 need to do
- ssh to root@ladd00
- edit /etc/auto.daq to add new machine...
- make -C /var/yp
Enable NFS V4 SERVER (SL6)
- if used with NIS, same as NFSv3
- if used as standalone, need to edit idmapd.conf - set the "Domain" name to the same value on NFS server and NFS slave (default automagically determined value does not always work). More TBW.
Enable AMANDA backups
AMANDA backups are already enabled by TRIUMF kickstart installs. For non-kickstart installation, follow instructions at [http://amanda/~amanda], or look at "/triumfcs/trshare/olchansk/linux/amanda/amanda-enable.perl". As final step, use [https://helpdesk.triumf.ca] to contact TRIUMF CS to add this new machine to the amanda backup list.
- yum install triumf-amanda
Enable AMANDA backups (CentOS7)
yum install amanda-client list-unit-files | grep -i amanda #systemctl enable amanda systemctl enable amanda.socket systemctl enable amanda-udp.socket systemctl restart amanda.socket systemctl restart amanda-udp.socket firewall-cmd --get-services firewall-cmd --permanent --add-service=amanda-client firewall-cmd --reload firewall-cmd --list-all echo amanda.triumf.ca amanda amdump >> /var/lib/amanda/.amandahosts
On amanda server, add new machine to the disklist, then:
amcheck -c daily titan00
Enable DCACHE
DAQ dcache server is mounted as
/daq/pnfs/triumf.ca/data/
For Centos-7 machines, you need to adjust the firewall rules in order to be able to communicate with the trdata machines; this is only necessary if you are copying data to trdata. The firewall changes are
firewall-cmd --permanent --add-rich-rule="rule family="ipv4" source address="142.90.100.212/32" port protocol="tcp" port="0-65535" accept" firewall-cmd --permanent --add-rich-rule="rule family="ipv4" source address="142.90.107.156/32" port protocol="tcp" port="0-65535" accept" firewall-cmd --permanent --add-rich-rule="rule family="ipv4" source address="142.90.100.219/32" port protocol="tcp" port="0-65535" accept" firewall-cmd --reload firewall-cmd --list-all
This instructions are unnecessary
- # mkdir -p /pnfs
- # edit /etc/rc.local, add to the end of file: "mount -o intr,rw,noac,hard,nfsvers=3 trdata00:/pnfs /pnfs &"
- # . /etc/rc.local
For more information on, see TrdataDcache dcache page.
Configure CPU speed (CentOS7)
In el7 the CPU frequency selection is confused. On some machines the default governor is "conservative", on other machines it is "powersave".
The current configuration can be seen by: "cpupower frequency-info -p"
The actual cpu frequency can be seen by "cat /proc/cpuinfo | grep -i mhz" and by "cpupower monitor" (run them under "watch -d -n1").
The linux kernel documentation says "powersave" will set CPU frequency to the minimum value, forever. But on some machines (i.e. daq06, daq14) it is easy to see that the CPU frequency actually changes according to the CPU load. This is explained in the documentation for the intel_pstate" driver.
On machines where CPU frequency seems always stuck at mimimum, try this:
- set the governor to "performance": cpupower frequency-set -g performance
- see if frequency now changes according to load (good) or is stuck at maximum (not so good, but ok)
- make it permanent by adding this command to /etc/rc.local - echo cpupower frequency-set -g performance >> /etc/rc.local
Configure Ganglia
SL6 Ganglia instructions (EPEL6 ganglia-3.7.2)
/bin/rm /etc/gmond.conf yum install "*gmond*" /bin/rm /etc/ganglia/conf.d/ganglia-triumf-daq.conf /bin/cp -v /dev/null /etc/ganglia/conf.d/multicpu.conf /bin/cp -v /dev/null /etc/ganglia/conf.d/netstats.pyconf /bin/cp -v /dev/null /etc/ganglia/conf.d/diskstat.pyconf /bin/cp -v /dev/null /etc/ganglia/conf.d/procstat.pyconf /bin/cp ~/git/scripts/etc/gmond.conf /etc/ganglia/gmond.conf chkconfig gmond on service gmond restart
Configure Ganglia (Centos7)
CentOS7 Ganglia instructions (EPEL7 ganglia-3.7.2)
/bin/rm /etc/gmond.conf yum -y install "ganglia-gmond*" /bin/cp -v /dev/null /etc/ganglia/conf.d/multicpu.conf # collects useless data /bin/cp -v /dev/null /etc/ganglia/conf.d/netstats.pyconf # spews errors into syslog /bin/cp -v /dev/null /etc/ganglia/conf.d/diskstat.pyconf # collects useless data /bin/cp -v /dev/null /etc/ganglia/conf.d/procstat.pyconf # do not create /tmp/gmond.conf /bin/cp ~/git/scripts/etc/gmond.conf /etc/ganglia/gmond.conf systemctl enable gmond systemctl restart gmond systemctl status gmond
Configure TRIUMF DAQ packages
(+CentOS7)
cd /etc/yum.repos.d wget http://daq.triumf.ca/~daqweb/yum/triumf-daq.repo
Install Konstantin's packages
(+CentOS7)
yum --disablerepo=\* --enablerepo=triumf-daq --skip-broken install diskscrub emailonreboot monitor_nfs "ganglia-*" triumf_nodeinfo
Install memtest and PXE boot
cd /boot
wget http://ladd00.triumf.ca/tftpboot/memtest86+-5.01.bin.gz
wget http://ladd00.triumf.ca/tftpboot/memtest86+-4.20.bin.gz
wget http://ladd00.triumf.ca/tftpboot/memtest86+-4.10
wget http://ladd00.triumf.ca/tftpboot/gpxe-1.0.1+-gpxe.lkrn
emacs -nw /boot/grub/grub.conf
title memtest86+-5.01
root (hd0,0)
kernel /boot/memtest86+-5.01.bin.gz
title memtest86+-4.20
root (hd0,0)
kernel /boot/memtest86+-4.20.bin.gz
title memtest86+-4.10
root (hd0,0)
kernel /boot/memtest86+-4.10
title pxeboot
root (hd0,0)
kernel /boot/gpxe-1.0.1+-gpxe.lkrn
Install node monitoring
(+CentOS7)
yum --disablerepo=\* --enablerepo=triumf-daq --skip-broken install triumf_nodeinfo /usr/sbin/sendnodeinfo.perl --config ladd00.triumf.ca:8600 emacs -nw /etc/nodeinfo /usr/sbin/sendnodeinfo.perl ladd00.triumf.ca:8600
Install gonodeinfo node monitoring
(+Ubuntu, +CentOS7)
go to https://bitbucket.org/dd1/gonodeinfo follow instructions:
yum -y install golang mkdir ~/git cd ~/git git clone https://bitbucket.org/dd1/gonodeinfo.git cd gonodeinfo git pull make make install # install gonodeinfo agent cd ~ # this is important
- edit /etc/gonodeinfo.conf
- change "Description", "Location", "User" and "Administrator" as appropriate (or delete them)
- change "Servers" to read: Servers: ladd00.triumf.ca:8601
- run gonodeinfo
- if error is "connection refused". go to the nodeinfo server to add this client to the access control list:
- on the gonodeinfo server: run gonodereceive -a daq13
- try gonodeinfo again, there should be no error
- on the gonodeinfo server: run gonodereport, look at the web pages, the new machine should be listed now
Install latest system updates
(+CentOS7)
yum update -y
Configure TRIUMF Printers
chkconfig cups off service cups stop yum install triumf-printers
Configure TRIUMF Printers (CentOS7)
systemctl stop cups systemctl disable cups echo "ServerName printers.triumf.ca" > /etc/cups/client.conf lpstat -a
Disable syslog spam (CentOS7)
Default el7 config is spamming the syslog with useless messages "systemd: Starting Session", etc. Disable this:
echo auditctl -e 0 >> /etc/rc.local echo /usr/bin/systemd-analyze set-log-level notice >> /etc/rc.local /etc/rc.local
Install basic system packages (CentOS7)
(if starting from minimal system, basic system packages required:)
yum install -y which psmisc redhat-lsb-core xorg-x11-xauth xterm emacs-nox rsync tcpdump strace nfs-utils sysstat iftop tcsh
yum install -y gcc gcc-c++ gdb glibc-static libstdc++-static zlib zlib-devel openssl-devel httpd-tools
Install packages needed for QUARTUS, ROOT, EPICS and MIDAS DAQ
(+CentOS7)
yum install --skip-broken giflib.x86_64 sysstat "libusb-devel*" unixODBC-devel postgresql-devel libxml2-devel libXpm-devel libgfortran git compat-readline43 "graphviz*" dcap "tigervnc*" telnet glibc"*" strace "fftw*" libpng "freetype*" xpdf "xemacs*" tkcvs xterm mutt "*-g77*" joe "libXmu*" dcap-devel gsl-devel pcre-devel h5py gd-devel xorg-x11-fonts"*" minicom xfig"*" perl-BSD-Resource "net-snmp-*" readline-static git-all nasm imake tcl-devel gv xorg-x11-twm expat-devel screen compat-readline5 ImageMagick ImageMagick-devel wget alacarte scipy numpy sympy nedit gnuplot php-cli php-domxml-php4-php5 php-gd php-fpdf php-cli kdebase cmake tcpdump sqlite sqlite-devel kdegraphics gdisk lsof gconf-editor iftop tk-devel mcelog kdm blt itcl lz4 bzip2 pbzip2 apr-devel apr-util-devel net-tools golang"*" --exclude golang-cover"*"hg"*" --exclude golang"*"hg"*" --exclude golang-pkg"*" --exclude golang-github"*" --exclude golang"*"git"*" mesa"*" xerces-c"*" diffuse clang i2c-tools texlive-revtex texlive-revtex4 kile kbibtex xrdp glibc.i686 gimp gimp-data-extras perl-GD"*" perl-Math"*" perl-Statistics-Basic cmake3 cmake3-gui extra-cmake-modules python2-pip x2go"*"
(do not install boost on 32-bit machines)
yum install --skip-broken "boost-*"
(packages for 32-bit software compilation on 64-bit machines. this is optional)
yum install --skip-broken giflib.i386 giflib.i686 compat-libf2c-34.i386 compat-libf2c-34.i686 mysql-devel.i686 openssl-devel.i686 unixODBC-devel.i686 libstdc++-devel.i386 libstdc++-devel.i686 "zlib-*.i686" "libXext-*.i686" "libXtst-*.i686" glibc-static.i686 freetype.i686 fontconfig.i686 libpng.i686 libXrender.i686 glibc-devel.i686 libX11-devel.i686 libXpm-devel.i686 libXft-devel.i686 mysql-devel.i686 dcap-devel.i686 gsl-devel.i686 pcre-devel.i686 fontconfig-devel.i686 freetype-devel.i686 libpng-devel.i686 libjpeg-devel.i686 libgfortran.i686 libxml2-devel.i686 gd-devel.i686 readline-devel.i686 ncurses-devel.i686 libXdmcp.i686 readline-static.i686 compat-readline5.i686
yum install boost-devel.i686
(separately install these packages - they collide with the big bunch above)
yum install rdesktop
yum reinstall urw-fonts
Install libraries for PHYSICA (CentOS7)
To run physica built on el6 from git sources on el7, do this:
(building physica on el7 is nort supported at this time)
(see more http://www.triumf.info/wiki/DAQwiki/index.php/PHYSICA)
yum -y install libX11.i686 gd.i686 libpng12.i686 readline.i686 compat-libf2c-34.i686
Install additional desktop environements (CentOS7)
# LXQT (from EPEL) yum -y install "lxqt*" # Cinnamon desktop (from EPEL) yum -y install cinnamon # KDE5 not available yet # MATE (from epel) yum -y groupinstall "MATE Desktop" yum -y install mate-common mate-icon-theme-faenza mate-netspeed mate-sensors-applet mate-themes-extras mate-utils yum -y erase ModemManager abrt abrt-libs abrt-gui-libs # XFCE4 (from EPEL) yum -y groupinstall xfce yum -y install "xfce*plugin" xfce4-about --exclude xfce4-hamster-plugin yum -y erase bash-completion
- make the MATE desktop as default
cd ~root/git/scripts/ git pull /bin/cp -v etc/lightdm_default_mate.conf /etc/lightdm/lightdm.conf.d/
- lighdm login manager (from EPEL)
yum install lightdm lightdm-kde lightdm-qt lightdm-qt5
- and switch from gdm to lighdm
systemctl disable gdm.service systemctl enable lightdm.service (systemctl stop gdm; systemctl restart lightdm) &
Make installation smaller (optional)
This is optional. Only do this if reducing the size of the OS image is very important.
yum erase "texlive*" "java*" "boost*" yum erase "xemacs*" yum erase "libstdc++-docs"
Install SMART scripts
(+CentOS7)
ln -sf ~/git/scripts/smart-status/smart-status.perl ~/
Install NTFS drivers
yum install ntfs-3g ntfsprogs (from EPEL)
Install HFS and HFS+ drivers (CentOS7)
yum --disablerepo=\* --enablerepo=elrepo install kmod-hfs kmod-hfsplus
Install Google Chrome web browser (64-bit SL6)
Google-chrome 27 is too old to using with recent MIDAS but it has working Flash:
rpm -vh --install https://daqshare.triumf.ca/~olchansk/google-chrome/google-chrome-stable-27.0.1453.110-202711.x86_64.rpm /bin/rm /etc/cron.daily/google-chrome yum-config-manager --disable google-chrome yum-config-manager --disable google-chrome-64 google-chrome
Chromium 38 works with current MIDAS. No Flash, no PDF viewer:
yum install -y policycoreutils-python rpm -vh --install https://daqshare.triumf.ca/~olchansk/google-chrome/chromium-browser-38.0.2125.111-1.el6.centos.x86_64.rpm chromium-browser
Install Google Chrome web browser (64-bit CentOS7)
/bin/cp ~/git/scripts/etc/google-chrome-64.repo /etc/yum.repos.d/ yum install google-chrome-stable
Enable monitoring of HTTPS certificates
On SL6, CentOS7:
yum install crypto-utils /etc/cron.daily/certwatch strace -f /etc/cron.daily/certwatch |& grep open | grep crt
Enable 100dpi fonts for EPICS
(+CentOS7)
ln -s /usr/share/X11/fonts/100dpi /etc/X11/fontpath.d/
Enable crontab @reboot for MIDAS (CentOS7)
el7 has a bug - cron @reboot entries for normal users can run before autofs is ready, so if the home directory is on autofs/NFS, it cannot be accessed and the cron job fails. If MIDAS is supposed to be started by cron @reboot, it will not start (there *will* be an error message in /var/log/cron).
mkdir /etc/systemd/system/crond.service.d echo -e "[Unit]\nAfter=ypbind.service autofs.service\n" > /etc/systemd/system/crond.service.d/local.conf systemctl daemon-reload systemctl cat crond.service
Explore the systemd dependacy tree using "systemctl list-dependencies" maybe with "--all".
Visualize the exact boot sequence from previous boot: "systemd-analyze plot > xxx.svg", look at the svg file using a web browser.
Enable firewall for MIDAS (CentOS7)
Default el7 configuration prevents all access to servers running on the local machine, including access to MIDAS mhttpd (tcp port 8443) and mserver (all tcp ports).
To enable access to mhttpd:
firewall-cmd --add-port=8443/tcp --permanent firewall-cmd --reload firewall-cmd --list-all
To enable access to the mserver from a specific host: (replace 142.90.111.175 with the IP address of the permitted host)
firewall-cmd --permanent --add-rich-rule="rule family="ipv4" source address="142.90.111.175/32" port protocol="tcp" port="0-65535" accept" firewall-cmd --reload firewall-cmd --list-all
Enable firewall for EPICS (CentOS7)
To enable access to TRIUMF EPICS servers, do this:
firewall-cmd --permanent --add-rich-rule="rule family="ipv4" source address="142.90.132.0/23" accept" firewall-cmd --reload firewall-cmd --list-all
For UCN the controls people seem to have EPICS setup on a different server; this might be true for CMMS as well. In this case the firewall rule change should be
firewall-cmd --permanent --add-rich-rule="rule family="ipv4" source address="142.90.139.0/23" accept" firewall-cmd --reload firewall-cmd --list-all
Disable gdm and X11 (OPTIONAL)
initctl stop prefdm echo "start on never" > /etc/init/prefdm.override echo "start on never" > /etc/init/splash-manager.override initctl reload-configuration
then enable login on default console:
echo "plymouth quit" >> /etc/rc.local echo "X_TTY=xxx/dev/tty1" >> /etc/sysconfig/init
Install JAVAWS (OPTIONAL)
- to run Java "web start" jnlp files (EVO, SEEVOGH, etc): javaws Downloads/spider.jnlp
- install javaws:
- yum install icedtea-web icedtea-web-javadoc
Install firefox java plugin (OPTIONAL, DO NOT DO THIS)
This installs the Oracle Java plugin:
- rpm -vh --install ~deap/jdk-7u15-linux-x64.rpm
- ls -l /usr/lib64/mozilla/plugins/
- ln -s /usr/java/jdk1.7.0_15/jre/lib/amd64/libnpjp2.so /usr/lib64/mozilla/plugins/
- start firefox, go edit->preferences->general->manage add-ons->plugins
- "java plugin 1.7.0_15" should be listed
Configure USB device permissions
(+CentOS7)
Configure USB device permissions for user access to USB-serial devices, Altera USB Blaster, etc.
- create file /etc/udev/rules.d/99-usb-chmod.rules with this contents:
emacs -nw /etc/udev/rules.d/99-usb-chmod.rules
ACTION=="add", SUBSYSTEM=="usb_device", RUN+="/bin/chmod a+wr /dev/%c"
ACTION=="add", SUBSYSTEM=="usb_device", RUN+="/bin/chmod a+wr /proc/%c"
ACTION=="add", ENV{DEVTYPE}=="usb_device", RUN+="/bin/chmod a+wr $env{DEVNAME}"
ACTION=="add", ENV{DEVTYPE}=="usb_device", RUN+="/bin/chmod a+wr $env{DEVICE}"
ACTION=="add", ENV{PHYSDEVBUS}=="usb-serial", RUN+="/bin/chmod a+wr $env{DEVNAME}"
ACTION=="add", ENV{DEVPATH}=="/class/tty/ttyS*", RUN+="/bin/chmod a+wr $env{DEVNAME}"
ACTION=="add", SUBSYSTEM=="tty", DEVPATH=="*ttyUSB*", RUN+="/bin/chmod a+rw $env{DEVNAME}"
ACTION=="add", SUBSYSTEM=="tty", DEVPATH=="*ttyS*", RUN+="/bin/chmod a+rw $env{DEVNAME}"
- apply new permissions: udevadm trigger --action=add
Disable modem-manager
The modem-manager will try to talk to any serial devices attached to USB serial ports. It assumes that those devices are modems and will send out modem-specific commands. if the devices are not modems and do not understand or do not like modem commands, well that's too bad. modem-manager is installed by the ModemManager package required by the NetworkManager package, and there is no configuration setting to turn modem-manager off.
One way to disable it is: chmod a= /usr/sbin/modem-manager
Another way to disable it is by forced uninstall: rpm --erase --nodeps ModemManager
Remember to kill the running copy: killall -KILL modem-manager
Caveat: it is not clear if modem-manager would not be resurrected by an update to the NetworkManager or ModemManager packages.
Configure Altera jtagd
(if needed)
mkdir /etc/jtagd echo 'Password = "123";' > /etc/jtagd/jtagd.conf cp -pv /triumfcs/trshare/olchansk/altera/11.0/quartus/linux/pgm_parts.txt /etc/jtagd/jtagd.pgm_parts
- start local jtagd: /triumfcs/trshare/olchansk/altera/11.0/quartus/bin/jtagd
- test local connection: /triumfcs/trshare/olchansk/altera/11.0/quartus/bin/jtagconfig
- test remote connection (add this machine to your .jtag.conf, run jtagconfig
For more information, go to Quartus
Install EOS
Instructions from here: http://eos-docs.web.cern.ch/eos-docs/quickstart/setup_repo.html
rpm -vh --install https://dss-ci-repo.web.cern.ch/dss-ci-repo/eos/citrine/tag/el-7/x86_64/eos-repo-el7-generic-1.noarch.rpm yum-config-manager --disable eos-citrine # disable auto-update because all packages are not signed yum-config-manager --disable eos-dep # disable auto-update because all packages are not signed. yum install eos-client eos-fuse --enablerepo=eos-citrine
Install fix for the el7 systemd dbus boot hang
Around early Summer 2018 el7 started showing a boot problem. In the nutshell, there is a problem with the dbus connection between dbus and systemd that prevents polkit, firewalld, etc from starting. The system eventually boots enough that one can ssh into it, but most things do not work. Notably, polkit is not running, firewalld is not running, ssh login takes about 15-30 second.
Solution is to add a special systemd service to check that dbus started correctly. It that runs after dbus is started, but before it is used, and it restarts dbus in a loop with a delay until dbus starts correctly. In testing, dbus always starts correctly after the first retry.
cd ~root/git/scripts/etc git pull /bin/cp -vf systemd-check-dbus.perl /usr/bin/ /bin/cp -vf systemd-check-dbus.service /etc/systemd/system/ systemctl daemon-reload systemctl enable systemd-check-dbus systemctl start systemd-check-dbus systemctl status systemd-check-dbus
After linux boots, if everything was okey, the script will report this:
[root@iris01 ~]# systemctl status systemd-check-dbus ... Feb 08 17:15:49 iris01.triumf.ca systemd[1]: Starting Check that systemd is registered with dbus... Feb 08 17:15:49 iris01.triumf.ca sh[4283]: Starting check for systemd dbus connection Feb 08 17:15:50 iris01.triumf.ca sh[4283]: List: string "org.freedesktop.DBus" Feb 08 17:15:50 iris01.triumf.ca sh[4283]: List: string "org.freedesktop.systemd1" Feb 08 17:15:50 iris01.triumf.ca sh[4283]: systemd1 dbus service exists, success! Feb 08 17:15:50 iris01.triumf.ca sh[4283]: Finished check for systemd dbus connection Feb 08 17:15:50 iris01.triumf.ca systemd[1]: Started Check that systemd is registered with dbus.
If the boot problem happened, the script will report about restarting dbus.
Note: the systemd service file adjusts the start order of other services, this adjustment seems to reduce the probability of the problem.
Configure GRUB boot loader
- edit /boot/grub/grub.conf, remove the "quiet" and "rhgb" options
- edit /boot/grub/grub.conf, comment out (with "#") the "splashimage=" line
- check that GRUB boot loader is installed on all system disks:
- dd if=/dev/sda bs=1 count=1024 2>&1 | strings | grep GRUB
- dd if=/dev/sdb bs=1 count=1024 2>&1 | strings | grep GRUB
- if GRUB is not installed, (i.e. on the 2nd disk of machines with mirrored system disk), (but check that /dev/sdb is the right disk):
# grub grub> device (hd0) /dev/sdb grub> root (hd0,0) grub> setup (hd0)
Configure GRUB boot loader (CENTOS7)
- edit /etc/default/grub, remove "rhgb" and "quiet" from GRUB_CMDLINE_LINUX
- grub2-mkconfig -o /boot/grub2/grub.cfg
- grub2-mkconfig -o /boot/efi/EFI/redhat/grub.cfg
- grub2-editenv list # show contents of boot environement file
- /bin/rm /boot/grub2/grubenv # remove stale settings, make grub2 boot from first entry in config file
Install memtest86+ (CentOS7)
yum -y install memtest86+ /bin/cp -vf /usr/share/memtest86+/20_memtest86+ /etc/grub.d/ /bin/chmod a+x /etc/grub.d/20_memtest86+ grub2-mkconfig -o /boot/grub2/grub.cfg
Configure GRUB boot loader (CentOS7)
DO NOT DO ANY OF THIS.
- (maybe) grub2-install /dev/sda
- check that GRUB boot loader is installed on all system disks:
- dd if=/dev/sda bs=1 count=1024 2>&1 | strings | grep GRUB
- dd if=/dev/sdb bs=1 count=1024 2>&1 | strings | grep GRUB
- if GRUB is not installed, (--- unfinished)
Disable ELREPO
sed 's/enabled=.*/enabled=0/' -i /etc/yum.repos.d/elrepo_triumf.repo sed 's/enabled=.*/enabled=0/' -i /etc/yum.repos.d/elrepo.repo
Special hardware settings
ASUS Crosshair mobo
- use BIOS version 1207 or newer
- (before CentOS7) sensors need these drivers from ELREPO: yum install --noplugins kmod-it87 kmod-k10temp; sensors-detect; service lm_sensors restart; sensors
- CentOS7: installs correct drivers automatically
ASUS Crosshair-II mobo
- use BIOS version 2607 or newer
- for the onboard IDE to work, add "all-generic-ide" to kernel boot options in grub.conf
- sensors need these drivers from ELREPO: yum install --noplugins kmod-it87 kmod-k10temp; sensors-detect; service lm_sensors restart; sensors
ASUS P7P55D EVO mobo
- use BIOS version 2004 or newer
- SL6 - install special driver for on board PCIe GigE network port and disable on board PCI GigE network port:
- yum --enablerepo elrepo install kmod-r8168 kmod-r8169
- # do not do this: sed 's/^blacklist/#blacklist/' -i /etc/modprobe.d/blacklist-r8169.conf
- reboot
- verify that correct drivers are loaded: ethtool -i eth0; ethtool -i eth1
- note: there will be no eth1 - r8169 driver is disabled.
ASUS P6X58-E-WS mobo
- BIOS settings
- F1 or DEL to enter BIOS setup, F8 boot menu
- go to POWER->HW mon, confirm CPU temperature is around 30C. (heatsink is installed correctly. Bad heatsink temperature quickly goes up to 50-70C).
- Main menu: Storage config - SATA change IDE->AHCI
- System information: confirm BIOS version 301, CPU type, memory size
- AI Tweak: set DRAM frequency - AUTO->DDR3-1333
- Advanced->Onboard devices: LAN BOOT: enabled
- Power->HW monitor: CPU Q-FAN: enabled
- Boot->Settings: Quick boot: enabled; Full screen logo: disabled; Wait for F1: disabled
- Save and exit
ASUS E35M1-M PRO mobo
- http://www.asus.com/Motherboards/E35M1M_PRO/#specifications
- use BIOS version 1002 or newer
- for CPU temperature: install kmod-k10temp from ELREPO (kmod-k10temp-0.0-4.el6.elrepo.x86_64.rpm)
- for Sensors: yum --enablerepo elrepo install kmod-w83627ehf; modprobe w83627ehf; sensors
- for Graphics: yum --enablerepo elrepo install kmod-fglrx fglrx-x11-drv
- to enable booting from USB3, edit /etc/dracut.conf, change line "add_drivers" to read: add_drivers+="xhci-hcd"
- to use multiple monitors, run "aticonfig --initial --heads=2 --adapter=1 --xinerama=on", to change screen layout, edit /etc/X11/xorg.conf. Only dual monitors DVI+HDMI seem to work. Tripple monitors does not seem to work.
Sensors instructions below are obolete (use driver from ELREPO)
- for Sensors, install driver for NCT6776F chip from https://github.com/groeck/w83627ehf/archives/master (in the Makefile, change the line "KERNEL_BUILD=" to read: "KERNEL_BUILD:=/usr/src/kernels/$(TARGET)"):
cd ~root wget http://ladd00.triumf.ca/~olchansk/linux/groeck-w83627ehf-dd3e543/w83627ehf.ko echo "modprobe hwmon; modprobe hwmon-vid; modprobe k10temp; rmmod w83627ehf; insmod /root/w83627ehf.ko" >> /etc/rc.local
ASUS E45M1-M PRO mobo
- https://www.asus.com/Motherboards/E45M1M_PRO/#specifications
- use BIOS 1202 or newer
- follow the E35M1-M PRO instructions above
ASUS P9X79 WS
- http://www.asus.com/Motherboard/P9X79_WS/
- use BIOS version 3101, 3401, 4701 or newer. If BIOS is 1305 or older, install P9X79-WS-CAP-Converter.ROM (BIOS 2902/3101), then the new BIOS.
- (not needed for CentOS7) for CPU temperature, install coretemp
- (not needed for CentOS7) for sensors, install driver for NCT6776F chip same as E35M1-M above.
- BIOS Settings:
- enter "Advanced mode"
- Ai Tweaker -> Ai Overclock Tuner -> Set to "XMP" - this enables DDR3-1600 RAM speed vs DDR3-1333 by default
- Monitor -> CPU fan speed low limit -> Set to "200 RPM" - we are using high efficiency slow turning CPU coolers and the default 600 RPM is right on the edge of firing false warnings
- Boot -> Full screen logo -> Set to "disabled"
- Wait for F1 -> Set to "disabled"
ASUS P8B-M
- use BIOS version 6103 or newer
- for CPU temperature, install coretemp
- for sensors, install driver for NCT6776F chip same as E35M1-M above.
SUPERMICRO X9SCL
- yum install kmod-w83627ehf.x86_64 coretemp
- xemacs -nw /etc/rc.local, add:
modprobe coretemp modprobe w83627ehf
ASUS Z87-WS
cd ~root wget http://ladd00.triumf.ca/~olchansk/linux/nct6775/nct6775.ko
Place the modprobe and insmod lines in /etc/rc.local to load the drivers at boot time
modprobe hwmon-vid insmod /root/nct6775.ko
ASUS AM1M-A
- use BIOS 602 or later
- SL6.5 installer cannot use USB2 ports and the network. Use USB3 ports (blue colour) to boot USB installer (memtest, rescue, etc)
- SL6.5 kernels require boot option "iommu=soft" or USB2 and network do not work. (USB3 - blue ports - seems okey)
- install ATI/AMD video drivers from ELREPO (see below)
- sensors chip is ITE IT8623E, for SL6, use standalone driver from lm_sensors. (2 fans rpm, 2 temperatures):
cd ~root wget http://ladd00.triumf.ca/~olchansk/linux/it87.ko echo modprobe hwmon_vid >> /etc/rc.local echo insmod /root/it87.ko >> /etc/rc.local . /etc/rc.local
- for el7 use it87.ko driver:
cd ~root wget https://daqshare.triumf.ca/~olchansk/linux/CentOS7/it87.ko echo modprobe hwmon_vid >> /etc/rc.local echo insmod /root/it87.ko >> /etc/rc.local . /etc/rc.local
- sensors output:
[root@midemma02 ~]# sensors
radeon-pci-0008
Adapter: PCI adapter
temp1: +22.0°C (crit = +120.0°C, hyst = +90.0°C)
fam15h_power-pci-00c4
Adapter: PCI adapter
power1: N/A (crit = 25.00 W)
k10temp-pci-00c3
Adapter: PCI adapter
temp1: +22.2°C (high = +70.0°C)
(crit = +70.0°C, hyst = +69.0°C)
it8603-isa-0290
Adapter: ISA adapter
in0: +0.96 V (min = +2.50 V, max = +2.95 V) ALARM
in1: +2.23 V (min = +0.94 V, max = +1.22 V) ALARM
in2: +2.03 V (min = +0.74 V, max = +0.77 V) ALARM
in3: +2.00 V (min = +1.26 V, max = +0.13 V) ALARM
in4: +2.23 V (min = +2.95 V, max = +2.15 V) ALARM
3VSB: +3.36 V (min = +6.00 V, max = +2.50 V) ALARM
Vbat: +3.22 V
+3.3V: +3.36 V
fan1: 611 RPM (min = 200 RPM)
fan2: 707 RPM (min = 600 RPM) ALARM
temp1: +38.0°C (low = +122.0°C, high = +122.0°C) sensor = thermistor
temp2: +22.0°C (low = +119.0°C, high = -35.0°C) ALARM sensor = thermistor
temp3: -128.0°C (low = +16.0°C, high = +93.0°C) sensor = thermistor
intrusion0: ALARM
[root@midemma02 ~]#
- AMD "Athlon(tm) 5350 APU" graphics supports 2 monitors maximum (mobo has 3 video outputs, only 2 can be used together)
Intel SE7230NH1
- front panel header connector pinout is like this:
PWR LED | 1 2|
| 3 4|
PWR LED | 5 6|
HDD LED | 7 8|
HDD LED | 9 10|
PWR SW |11 12| NIC1 LED
PWR SW |13 14| NIC1 LED
RST SW |15 16|
RST SW |17 18|
|19 20|
NMI SW |21 22| NIC2 LED
NMI SW |23 24| NIC2 LED
... |... |
|33 34|
ASUS H110M-A/M.2
- use BIOS 2003 or later
- sensors chip is ??? for el7, use this driver:
cd ~root wget https://daqshare.triumf.ca/~olchansk/linux/CentOS7/nct6775.ko echo modprobe hwmon_vid >> /etc/rc.local echo modprobe coretemp >> /etc/rc.local echo insmod /root/nct6775.ko >> /etc/rc.local . /etc/rc.local
- sensors output:
[root@daq03 ~]# sensors
acpitz-virtual-0
Adapter: Virtual device
temp1: +27.8°C (crit = +119.0°C)
temp2: +29.8°C (crit = +119.0°C)
nct6793-isa-0290
Adapter: ISA adapter
in0: +0.34 V (min = +0.00 V, max = +1.74 V)
in1: +1.02 V (min = +0.00 V, max = +0.00 V) ALARM
in2: +3.39 V (min = +0.00 V, max = +0.00 V) ALARM
in3: +3.39 V (min = +0.00 V, max = +0.00 V) ALARM
in4: +1.02 V (min = +0.00 V, max = +0.00 V) ALARM
in5: +0.15 V (min = +0.00 V, max = +0.00 V) ALARM
in6: +0.97 V (min = +0.00 V, max = +0.00 V) ALARM
in7: +3.38 V (min = +0.00 V, max = +0.00 V) ALARM
in8: +3.12 V (min = +0.00 V, max = +0.00 V) ALARM
in9: +1.00 V (min = +0.00 V, max = +0.00 V) ALARM
in10: +0.14 V (min = +0.00 V, max = +0.00 V) ALARM
in11: +0.12 V (min = +0.00 V, max = +0.00 V) ALARM
in12: +0.14 V (min = +0.00 V, max = +0.00 V) ALARM
in13: +0.12 V (min = +0.00 V, max = +0.00 V) ALARM
in14: +0.13 V (min = +0.00 V, max = +0.00 V) ALARM
fan1: 1041 RPM (min = 0 RPM)
fan2: 1020 RPM (min = 0 RPM)
fan5: 0 RPM (min = 0 RPM)
fan6: 0 RPM
SYSTIN: +119.0°C (high = +98.0°C, hyst = +95.0°C) sensor = thermistor
CPUTIN: +26.5°C (high = +80.0°C, hyst = +75.0°C) sensor = thermistor
AUXTIN0: +27.5°C sensor = thermistor
AUXTIN1: +112.0°C sensor = thermistor
AUXTIN2: +111.0°C sensor = thermistor
AUXTIN3: +111.0°C sensor = thermistor
PECI Agent 0: +28.0°C (high = +98.0°C, hyst = +95.0°C)
(crit = +100.0°C)
PECI Agent 0 Calibration: +25.5°C
PCH_CHIP_CPU_MAX_TEMP: +0.0°C
PCH_CHIP_TEMP: +0.0°C
intrusion0: ALARM
intrusion1: ALARM
beep_enable: disabled
coretemp-isa-0000
Adapter: ISA adapter
Physical id 0: +31.0°C (high = +80.0°C, crit = +100.0°C)
Core 0: +31.0°C (high = +80.0°C, crit = +100.0°C)
Core 1: +28.0°C (high = +80.0°C, crit = +100.0°C)
[root@daq03 ~]#
Supermicro X11SSH-F
- blacklist the mei and mei_me drivers per http://www.supermicro.com/support/faqs/faq.cfm?faq=14537
[root@alpha00 ~]# more /etc/modprobe.d/blacklist.conf blacklist mei blacklist mei_me [root@alpha00 ~]#
- mobo requires M.2 PCIe SSD (M.2 SATA SSD would not work. SATA SATA SSD ok)
- boot from M.2 PCIe SSD requires UEFI boot (from an MSDOS partition on the SSD)
Configure X11 graphics
Special settings for DAQ
- add the following at the end of /etc/X11/xorg.conf. The enables Ctrl-Alt-KP-/ and Ctrl-Alt-KP-* to unlock the keyboard after Altera Quartus crash:
Section "ServerFlags" Option "AllowDeactivateGrabs" "true" Option "AllowClosedownGrabs" "true" EndSection
Install NVIDIA drivers
- yum --enablerepo=elrepo install nvidia-detect
- run: nvidia-detect
- as instructed by nvidia-detect, install correct driver:
- yum --enablerepo=elrepo install kmod-nvidia
- yum --enablerepo=elrepo install kmod-nvidia-304xx
- yum --enablerepo=elrepo install kmod-nvidia-173xx
- (before SL6.x: if it fails due to conflict with module-init-tools, run "yum --disablerepo \* --enablerepo elrepo update module-init-tools")
- yum erase xorg-x11-glamor ### see http://elrepo.org/tiki/kmod-nvidia (search for glamor)
- mv /etc/X11/xorg.conf /etc/X11/xorg.conf-xxx
- nvidia-xconfig
- (SL6) reboot
- (SL5) /dev/MAKEDEV nvidia
- (SL5) restart the X11 server (Ctrl-Alt-Backspace or "killall Xorg gdm-binary")
- observe that X11 server restarts using the NVIDIA driver (big NVIDIA logo on startup)
- if needed, login as root and run "nvidia-settings" to setup dual-screen configuration, etc
Install legacy NVIDIA drivers
For old NVIDIA cards:
- GeForce FX 5500
wget http://us.download.nvidia.com/XFree86/Linux-x86/173.14.31/NVIDIA-Linux-x86-173.14.31-pkg1.run sh ./NVIDIA-Linux-x86-173.14.31-pkg1.run
- GeForce 6200 - NVIDIA Corporation NV44A [GeForce 6200]
yum install nvidia-x11-drv-304xx-304.121 --enablerepo=elrepo nvidia-xconfig rmmod nvidia killall gdm-binary login as root nvidia-settings to setup multiple displays
Install ATI/AMD drivers
- yum --enablerepo elrepo install kmod-fglrx fglrx-x11-drv
- check that /etc/X11/xorg.conf section "Device" entry "Driver" says "fglrx"
- run "aticonfig --initial" to create xorg.conf if existing one is not good
- run "amdcccle" as root to configure dual-screens, etc
Note: 'amdcccle' is a GUI, so you must run this command from within a running X session
- killall Xorg
Install ATI/AMD drivers (CentOS7)
- wget http://elrepo.org/linux/testing/el7/x86_64/RPMS/fglrx-x11-drv-15.12-3.el7.elrepo.x86_64.rpm
- wget http://elrepo.org/linux/testing/el7/x86_64/RPMS/kmod-fglrx-15.12-3.el7.elrepo.x86_64.rpm
- yum install acpid
- rpm -vh --install kmod-fglrx-15.12-3.el7.elrepo.x86_64.rpm fglrx-x11-drv-15.12-3.el7.elrepo.x86_64.rpm
- amdconfig -f --initial
- grub2-mkconfig -o /boot/grub2/grub.cfg
- reboot
- login as root
- amdcccle
NOTE: if both drivers - radeon and fglrx are loaded, boot will hang. the radeon driver is supposed to be blacklisted through grub rdblacklist=radeon entry which is installed by running grub2-mkconfig.
Install Intel drivers for HD4600/Z87
SL6.5 has the required drivers for the socket 1150 machines with Intel HD4600 graphics and Z87 chipset.
ASUS Z87 WS motherboard has these video connections with corresponding Intel video port assignements, as reported by "xrandr":
- DisplayPort - DP1/HDMI1
- MiniDisplayPort - DP2/HDMI2
- HDMI - HDMI3
Due to hardware limitations, 3 HDMI monitors using 2 passive DP-HDMI adapters (and 1 straight HDMI) cannot be used.
To use 3 monitors do this:
- 1st monitor: DisplayPort - DP-to-HDMI-passive-adapter - HDMI monitor (not tried: DP-to-DP-cable - DisplayPort monitor).
- 2nd monitor: MiniDisplayPort - MiniDP-to-DP-cable - DisplayPort monitor
- 3rd monitor: HDMI - HDMI-cable - HDMI monitor
With the monitors I have (Dell 1920x1200 VGA-HDMI-DP), the software thinks that there are 4 monitors: somehow both DP2 and HDMI2 see 1 minitor each, but the hardware cannot drive 4 monitors, so everything goes blank. To fix, disable HDMI2 (xrandr -display :0 --output HDMI2 --off) and enable DP2 (xrandr -display :0 --output DP2 --auto).
How to make this configuration permanent and how to assign monitor locations (left-right, etc), you figure it out.
Manual selection of monitor, video mode and resolution
Automatic selection of monitor and video mode usually works. When it does not, configure it manualls:
- physically go to the computer
- login as root
- run "nvidia-settings" on machines using the NVIDIA driver
- run "aticonfig" on machines with the ATI/AMD driver (use "aticonfig --initial" for initial setup, and good luck with anything more complicated)
- run "system-config-display".
- In the "hardware" tab, select monitor type: "generic LCD 1280x1024" or "generic LCD 1600x1200".
- In the "settings" tab, select "1280x1024" or "1600x1200" and "Thousands of colors".
- Press "ok", the display settings application should close.
- Logout, the new login window should use the new settings.
Disable screen saver
If machine is booted without any monitor connected, current video cards to not enable any video outputs. If a monitor is connected later, there is no video image and there is no easy way to get a video image.
This can be solved by configuring X11 to always enable some video output. Because the monitor type is not known when X11 starts, one has to select some standard video mode (i.e. VESA 1280x1024) on some video output (VGA, DVI or HDMI).
Only NVIDIA cards with the NVIDIA driver (from EPEL) is supported by these instructions.
- create default xorg.conf: nvidia-xconfig
- edit /etc/X11/xorg.conf
- add monitor section for the fake monitor:
Section "Monitor"
Identifier "Monitor0"
VendorName "Unknown"
ModelName "Unknown"
HorizSync 31.0 - 83.0
VertRefresh 59.0 - 61.0
Option "DPMS" "off"
ModeLine "1280x1024" 108.00 1280 1328 1440 1688 1024 1025 1028 1066 +hsync +vsync
EndSection
- add output selection in the "Device" section:
Section "Device"
Identifier "Device0"
Driver "nvidia"
VendorName "NVIDIA Corporation"
BoardName "GeForce 210"
#Option "ConnectedMonitor" "DFP"
#Option "ConnectedMonitor" "CRT"
Option "ConnectedMonitor" "CRT-1"
Option "UseEDID" "no"
EndSection
- add fake video mode to the "Screen" section:
Section "Screen"
Identifier "Screen0"
Device "Device0"
Monitor "Monitor0"
DefaultDepth 24
SubSection "Display"
Depth 24
Modes "1280x1024"
EndSubSection
EndSection
- disable screen saver and DPMS power off in the "ServerLayout" or "ServerFlags" section:
Section "ServerLayout"
Identifier "Layout0"
Screen 0 "Screen0" 0 0
InputDevice "Keyboard0" "CoreKeyboard"
InputDevice "Mouse0" "CorePointer"
Option "Xinerama" "0"
Option "BlankTime" "0"
Option "StandbyTime" "0"
Option "SuspendTime" "0"
Option "OffTime" "0"
EndSection
Section "ServerFlags"
Option "BlankTime" "0"
Option "StandbyTime" "0"
Option "SuspendTime" "0"
Option "OffTime" "0"
EndSection
Finish installation
- logout and reboot the computer to have all the changes to take effect
Configure HTTPS server (CentOS7)
This will configure the HTTPS/SSL certificate using "certbot" and "letsencrypt" and configure an HTTPS web server using apache httpd.
First, configure apache httpd:
- yum install mod_ssl certwatch crypto-utils
- cd /etc/httpd/conf.d/
- mv ssl.conf ssl.conf-not-used ### remove the stock ssl.conf which refers to the localhost certificate that will expire in 1 year
- touch ssl.conf ### create a blank file to prevent automatic updates from installing a stock ssl.conf file
- rm /etc/pki/tls/certs/localhost.crt
- create new file ssl-daq12.conf # use actual hostname instead of daq12
Listen 443 https #SSLPassPhraseDialog exec:/usr/libexec/httpd-ssl-pass-dialog SSLSessionCache shmcb:/run/httpd/sslcache(512000) SSLSessionCacheTimeout 300 SSLRandomSeed startup file:/dev/urandom 256 SSLRandomSeed connect builtin SSLCryptoDevice builtin <VirtualHost *:443> ServerName daq12.triumf.ca DocumentRoot /var/www/html ErrorLog /var/log/httpd/daq12.log SSLEngine on # note SSLProtocol, SSLCipherSuite and some other settings are overwritten by /etc/letsencrypt/options-ssl-apache.conf SSLProtocol all -SSLv2 -SSLv3 SSLCipherSuite HIGH:MEDIUM:!aNULL:!MD5:!SEED:!IDEA:!RC4 SSLCertificateFile /etc/pki/tls/certs/localhost.crt SSLCertificateKeyFile /etc/pki/tls/private/localhost.key #SSLCertificateChainFile /etc/pki/tls/certs/server-chain.crt #ProxyPass /elog/ http://localhost:8082/ retry=1 #ProxyPass / http://localhost:8080/ retry=1 Header always set Strict-Transport-Security "max-age=31536000; includeSubDomains" <Location /> SSLRequireSSL AuthType Basic AuthName "DAQ password protected site" Require valid-user # create password file: touch /etc/httpd/htpasswd # to add new user or change password: htpasswd /etc/httpd/htpasswd username AuthUserFile /etc/httpd/htpasswd </Location> </VirtualHost>
- stop httpd from listening on port 80: edit /etc/httpd/conf/httpd.conf, comment-out the line "Listen 80"
- systemctl enable httpd
- systemctl restart httpd
- systemctl status httpd
- try to access https://daq12.triumf.ca
- you should see a complaint about self-signed certificate
- you should see a request for password (do not login yet)
- if you get "connection refused", HTTPS port 443 may need to be enabled in the local firewall, then try again:
firewall-cmd --add-port=443/tcp --permanent firewall-cmd --reload firewall-cmd --list-all
Second, configure certbot:
(Note: as of 2018-01-18 certbot requires use of http port 80 to get the initial https certificate, renewal can continue to use the https port 443)
(Note: as of 2019-01-?? certbot requires use of port 80 for renewals)
- check that port 80 is not used by anything:
- netstat -an | grep LISTEN | grep ^tcp | grep 80
- lsof -P | grep -i tcp | grep LISTEN | grep 80
- if lsof reports that httpd is listening on port 80, follow the httpd instructions above (remove "listen 80" from httpd.conf
- yum install certbot python2-certbot-apache # (from EPEL)
- firewall-cmd --add-port=80/tcp --permanent
- firewall-cmd --reload
- firewall-cmd --list-all
- certbot certonly --standalone --installer apache # then answer questions:
- "activate HTTPS for daq12.triumf.ca" - say ok
- "enter email address" - enter your own email address
- "please read terms..." - read the terms and say "agree"
- it will take a few moments...
- "please choose..." - say "easy" (http access is disabled (a) by firewall, (b) by local configuration
- "congratulations..." - say ok.
- certbot install --apache --cert-name daq12.triumf.ca # then answer questions:
- "choose redirect..." - say "1" (no redirect)
- look inside ssl-daq12.conf to see that SSLCertificateFile & co point to certbot certificates in /etc/letsencrypt/live/daq12.triumf.ca/
- enable automatic renewal
systemctl enable certbot-renew.timer systemctl start certbot-renew.timer systemctl list-timers --all
- to check corrent renewal and to update the certbot config file in /etc/letsencrypt/renewal, run this:
- certbot renew --standalone --installer apache --force-renewal
NOTE: this certificate will expire in 3 months, automatic renewal should work starting with certbot-0.12.0-4.el7.noarch. Certificate expiration should be automatically detected by "certwatch" and email will be sent to local root user, to be forwarded to an actual person by ~root/.forward.
Third, activate password protection:
- as shown in the config file above, create password file and initial user: (replace "midas" with specific username)
touch /etc/httpd/htpasswd htpasswd /etc/httpd/htpasswd midas
Final test:
- access https://daq12.triumf.ca - https status should be "green"
- login with password should work
- the apache httpd test page should load
- check site security using the SSLlabs https tester. (I get grade "A-"): https://www.ssllabs.com/ssltest/
From here:
- enable proxy for MIDAS mhttpd - uncomment redirect in the config file above
- enable proxy for ELOG - ditto
- setsebool -P httpd_can_network_connect 1
- systemctl restart httpd
NOTE: if certbot fails with errors about 'module' object has no attribute 'pyopenssl', try this: pip install requests==2.6.0
Configure large RAID6 arrays
- connect the disks
- check the disks health
- run smart-status.perl
- partition the disks
- yum install gdisk
- gdisk /dev/sdX
- delete all partitions: o
- create new partition: n, enter, enter, enter, fd00 (default sizes, partition type fd00)
- write and exit: w
- check presence of all partitions:
- /bin/ls -l /dev/sd*1
- prepare to use an external bitmap file
- touch /md6bitmap
- edit /etc/fstab, change entry for root filesystem from: "defaults 1 1" to "defaults 0 0"
- edit /boot/grub/grub.conf, change entry "kernel ... ro ..." to "kernel ... rw ..."
- create raid array:
- mdadm --create /dev/md6 --level=6 --bitmap=/md6bitmap --raid-devices=10 /dev/sd[b-k]1
- mdadm -Ds >> /etc/mdadm.conf
- cleanup /etc/mdadm.conf
- echo "echo 16384 > /sys/block/md6/md/stripe_cache_size" >> /etc/rc.local
- echo "echo 1 > /sys/block/md6/md/sync_speed_min" >> /etc/rc.local
- source /etc/rc.local
- observe raid array rebuild:
- watch -d -n1 "cat /proc/mdstat"
Configure ZFS
Install ZFS
(from here: https://github.com/zfsonlinux/zfs/wiki/RHEL-%26-CentOS)
Follow the instructions for "kABI-tracking kmod" - dkms modules seem to always mess up the system when upgrading to next release of zfs.
#rpm -vh --install http://archive.zfsonlinux.org/epel/zfs-release.el7.noarch.rpm #yum install http://download.zfsonlinux.org/epel/zfs-release.el7.noarch.rpm #yum install http://download.zfsonlinux.org/epel/zfs-release.el7_3.noarch.rpm #yum install http://download.zfsonlinux.org/epel/zfs-release.el7_4.noarch.rpm #yum install http://download.zfsonlinux.org/epel/zfs-release.el7_5.noarch.rpm yum install http://download.zfsonlinux.org/epel/zfs-release.el7_6.noarch.rpm yum-config-manager --disable zfs yum-config-manager --enable zfs-kmod yum install zfs #sed 's/^SELINUX=.*/SELINUX=disabled/' -i /etc/selinux/config echo USE_DISK_BY_ID=\'yes\' >> /etc/default/zfs #shutdown -r now # required to load the zfs kernel modules and to disable selinux modprobe zfs # should work zpool status # should report no pools available
- Note: zfs and selinux and not compatible: with selinux enabled, files on zfs cannot be deleted (files are gone, but "df" does not go down, zfs-0.6.5.7-1.el7.centos.x86_64), see #https://github.com/zfsonlinux/zfs/issues/4845
- http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/zfs-quickstart.html)
- http://www.freebsd.org/cgi/man.cgi?query=zpool&sektion=8
Lock kernel and zfs packages
yum versionlock kernel yum versionlock zfs
Misc commands
- zpool status
- zpool get all
- zpool iostat 1
- zpool iostat -v 1
- zpool history
- zpool scrub data14
- zpool events
- arcstat.py 1
- cat /proc/spl/kstat/zfs/arcstats
- echo 30000000000 > /sys/module/zfs/parameters/zfs_arc_meta_limit
- echo 32000000000 > /sys/module/zfs/parameters/zfs_arc_max
- zfs get all
- zfs set dedup=verify zssd/nfsroot
- zpool create data14 raidz2 /dev/sd[b-h]1
- zfs create z8tb/data
- zfs destroy z8tb/data
- zpool add z10tb cache /dev/disk/by-id/ata-ADATA_SP550_2F4320041688
- parted /dev/sdx mklabel GPT
- blkid
- zpool iostat -v -q 1
- watch -d -n 1 "cat /proc/spl/kstat/zfs/arcstats | grep l2"
- zfs set primarycache=metadata tank/datab
- zfs set secondarycache=metadata tank/datab
- zfs userspace -p -H zssd/home1
- zfs groupspace ...
Create raid0 (mirror) volume
echo USE_DISK_BY_ID=\'yes\' >> /etc/default/zfs dracut -vf zpool create zssd mirror /dev/sdaX /dev/sdbX zpool set cachefile=none zssd zpool set failmode=continue zssd zpool status zpool events zpool get all df /zssd ls -l /zssd
Use whole disk for zfs mirror (RAID0)
echo USE_DISK_BY_ID=\'yes\' >> /etc/default/zfs
[root@daq13 ~]# parted /dev/sdb
(parted) mklabel GPT
(parted) q
[root@daq13 ~]# parted /dev/sdc
(parted) mklabel GPT
(parted) q
[root@daq13 ~]# blkid
/dev/sda1: UUID="ab920e4b-40ae-4551-aab8-f3e893d38830" TYPE="xfs"
/dev/sdb: PTTYPE="gpt"
/dev/sdc: PTTYPE="gpt"
[root@daq13 ~]# zpool create z10tb mirror /dev/sdb /dev/sdc
[root@daq13 ~]# zpool status
pool: z10tb
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
z10tb ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
sdb ONLINE 0 0 0
sdc ONLINE 0 0 0
errors: No known data errors
[root@daq13 ~]#
[root@daq13 ~]# zfs create z10tb/emma
[root@daq13 ~]# df -kl
Filesystem 1K-blocks Used Available Use% Mounted on
pool 9426697856 0 9426697856 0% /pool
pool/daqstore 9426697856 0 9426697856 0% /pool/daqstore
[root@daq13 ~]#
Enable ZFS at boot
systemctl enable zfs-import-cache systemctl enable zfs-import-scan systemctl enable zfs-mount systemctl enable zfs-import.target systemctl enable zfs.target
Replace failed disk
- pull failed disk out
- zpool status # identify failed disk zfs label (it should be labeled FAULTED or OFFLINE
- safe to reboot here
- install new disk
- partition new disk, i.e. "gdisk /dev/sdh", use "o" to create new partition table, use "n" to create new partition, accept all default answers, use "w" to save and exit
- safe to reboot here
- run tests on new disk (smart, diskscrub), if unhappy go back to "install new disk"
- safe to reboot here
- identify serial number of new disk, i.e. "smartctl -a /dev/sdh | grep -i serial" yields "Serial Number: WD-WCAVY0893313"
- identify linux id of new disk by "ls -l /dev/disk/by-id | grep -i WD-WCAVY0893313" yields "ata-WDC_WD2002FYPS-01U1B0_WD-WCAVY0893313-part1"
- zpool replace data11 zfs-label-of-failed-disk ata-WDC_WD2002FYPS-01U1B0_WD-WCAVY0893313-part1
- zpool status should look like this:
[root@daq11 ~]# zpool status
pool: data11
state: DEGRADED
status: One or more devices is currently being resilvered. The pool will
continue to function, possibly in a degraded state.
action: Wait for the resilver to complete.
scan: resilver in progress since Fri Apr 29 11:51:03 2016
24.7G scanned out of 795G at 32.3M/s, 6h46m to go
3.00G resilvered, 3.11% done
config:
NAME STATE READ WRITE CKSUM
data11 DEGRADED 0 0 0
raidz2-0 DEGRADED 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WCAZA3872943-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WCAZA1973466-part1 ONLINE 0 0 0
replacing-2 DEGRADED 0 0 0
17494865033746374811 FAULTED 0 0 0 was /dev/sdi1
ata-WDC_WD2002FYPS-01U1B0_WD-WCAVY0893313-part1 ONLINE 0 0 0 (resilvering)
ata-WDC_WD20EARS-00MVWB0_WD-WCAZA1973369-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WMAZA0858733-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WMAZA0819555-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WMAZA0857075-part1 ONLINE 0 0 0
ata-WDC_WD2002FYPS-01U1B0_WD-WCAVY0347413-part1 ONLINE 0 0 0
errors: No known data errors
- wait for raid rebuild ("resilvering") to complete
- zpool status should look like this:
[root@daq11 ~]# zpool status
pool: data11
state: ONLINE
scan: resilvered 96.2G in 1h44m with 0 errors on Fri Apr 29 13:35:40 2016
config:
NAME STATE READ WRITE CKSUM
data11 ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WCAZA3872943-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WCAZA1973466-part1 ONLINE 0 0 0
ata-WDC_WD2002FYPS-01U1B0_WD-WCAVY0893313-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WCAZA1973369-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WMAZA0858733-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WMAZA0819555-part1 ONLINE 0 0 0
ata-WDC_WD20EARS-00MVWB0_WD-WMAZA0857075-part1 ONLINE 0 0 0
ata-WDC_WD2002FYPS-01U1B0_WD-WCAVY0347413-part1 ONLINE 0 0 0
errors: No known data errors
Rename zfs pool
zpool export oldname zpool import oldname z6tb
Misc
ZFS tunable parameters for hopefully speeding up resilvering: https://www.reddit.com/r/zfs/comments/4192js/resilvering_raidz_why_so_incredibly_slow/ echo 0 > /sys/module/zfs/parameters/zfs_resilver_delay echo 512 > /sys/module/zfs/parameters/zfs_top_maxinflight echo 5000 > /sys/module/zfs/parameters/zfs_resilver_min_time_ms
Enable periodic scrub:
cd ~/git/scripts git pull cd zfs make install
Working with ZFS snapshots:
- zfs list -t snapshot
- cd ~/git; git clone https://github.com/zfsonlinux/zfs-auto-snapshot.git; cd zfs-auto-snapshot; make install
If ZFS becomes 100% full, "rm" will stop working, but space can still be freed by using "echo > bigfile", afterwards "rm" works again.
performance notes
Go here: disk_benchmarks
Configure UEFI boot
Some mobo can boot from NVME (PCIe) SSDs only via UEFI boot. Do this:
- partition the NVME SSD using gdisk (must be GPT partition table, must have MSDOS EFI partition size 512MiB)
[root@alpha00 ~]# gdisk -l /dev/nvme0n1 GPT fdisk (gdisk) version 0.8.6 ... Found valid GPT with protective MBR; using GPT. Disk /dev/nvme0n1: 500118192 sectors, 238.5 GiB Logical sector size: 512 bytes Disk identifier (GUID): 1A82CC87-2757-44ED-980F-C78E3681D9D3 Partition table holds up to 128 entries First usable sector is 34, last usable sector is 500118158 Partitions will be aligned on 2048-sector boundaries Total free space is 2014 sectors (1007.0 KiB) Number Start (sector) End (sector) Size Code Name 1 2048 1050623 512.0 MiB EF00 EFI System 2 1050624 500118158 238.0 GiB 8300 Linux filesystem [root@alpha00 ~]#
- create filesystems
mkfs.msdos /dev/nvme0n1p1 mkfs.xfs /dev/nvme0n1p2
- prepare EFI partition
mkdir /mnt/efi mount /dev/nvme0n1p1 /mnt/efi mkdir -p /mnt/efi/efi/boot cp /boot/vmlinuz... vmlinuz # copy the desired linux kernel cp /boot/initramfs... initramfs.img # copy the matching initramfs file #from /home/olchansk/sysadm/syslinux/syslinux-6.03 copy cp .../efi64/efi/syslinux.efi . cp .../efi64/com32/elflink/ldlinux/ldlinux.e64 . cp syslinux.efi bootx64.efi
- create syslinux config file: syslinux.cfg
default linux label linux kernel vmlinuz append ro root=/dev/nvme0n1p2 nomodeset initrd=initramfs.img
- prepare system partition
mkdir /mnt/tmp mount /dev/nvme0n1p2 /mnt/tmp rsync -avx / /mnt/tmp cd /mnt/tmp #edit etc/fstab #edit etc/syslinux/selinux # set selinux to permissive mode because rsync did not copy the selinux labels
- unmount and reboot
- restore selinux labels after first boot
#login as root cd / restorecon -R / # can also add "-v" to see progress, but runs much slower #edit /etc/sysconfig/selinux # enable selinux #shutdown -r now # reboot with selinux enabled