DEAP: Difference between revisions

Revision as of 09:14, 2 November 2012

DEAP DAQ machines are on the private network 192.168.1.x (netmask 255.255.255.0)

Gateway to TRIUMF network is 1U machine deap06.triumf.ca connected to the LADD-NIS cluster (deap account on ladd00).

Gateway services running on the gateway:

DHCP server for the 192.168.1.x network (/etc/hosts, /etc/dhcp/dhcpd.conf)
apache SSL/https proxy for MIDAS status page, ELOG, ganglia and nodeinfo (/etc/httpd/conf.d/ssl.conf, /etc/httpd/htpasswd)
NAT proxy from private network to the TRIUMF network (/etc/rc.local). Makes the internet accessible from deapNN machines.

The DEAP DAQ cluster is configured for standalone running with or without an internet connection.

(NB: Some internet functions are required: access to NTP for time synchronization and access to Linux package repositories to install packages, etc)

Usernames, passwords and hostnames are distributed using NIS:

domain name: DEAP-NIS
deap00 is the master server
there are no secondary servers
hostnames are distributed using NIS (from deap00:/etc/hosts, MUST MATCH deap06:/etc/hosts!)
to solve chicken-and-egg problem deap00 IP address has to be listed in each machine /etc/hosts (MUST MATCH deap06 and deap00 /etc/hosts!) (SL6.2+ NIS broadcast does not work so deap00 has to be listed in each machine /etc/yp.conf, also NFS filesystems are mounted before NIS is started).
also NIS has to be listed in front of DNS in the "hosts:" entry of /etc/nsswitch.conf

DNS kludge:

normally DNS would be used to distribute IP addresses and hostnames to the DHCP server, to deap00 and to other deap machines. But we do not have a private DNS server and the TRIUMF DNS server has the wrong IP addresses for deap machines (142.90.x.x).
deap06 DHCP is telling all machines to use the TRIUMF DNS server (to resolve internet addresses - google, etc). To avoid confusion between local deap00, etc hostnames and deap00, etc hostnames from TRIUMF, /etc/nsswitch.conf "hosts:" entry has to list "nis" before "dns".
hopefully the deap00, etc hostnames will be resolved correctly by the SNOlab DNS servers and all this kludging can go away.

System monitoring tools:

ganglia
triumf_nodeinfo
konstantin's ganglia packages (monitor_nfs, ganglia sensors, top, etc) - To install/update: yum --disablerepo="*" --enablerepo=konstantin update
diskscrub

backups of Linux images:
- backups of linux images are done to deap00:/data/root/backups using cron job on deap00:/etc/cron.d/backup.lxdaq.cron and deap00:~root/backup.lxdaq
backups of home directories: NONE
backups of data disks: NONE

@@ Line 22: / Line 22: @@
 Gateway services running on the gateway:
 * DHCP server for the 192.168.1.x network (/etc/hosts, /etc/dhcp/dhcpd.conf)
-* apache SSL/https proxy for MIDAS status page, ELOG, ganglia and nodeinfo (/etc/httpd/conf.d/ssl.conf)
+* apache SSL/https proxy for MIDAS status page, ELOG, ganglia and nodeinfo (/etc/httpd/conf.d/ssl.conf, /etc/httpd/htpasswd)
 * NAT proxy from private network to the TRIUMF network (/etc/rc.local). Makes the internet accessible from deapNN machines.