> Dear MIDAS experts,
> 
> when I attempted to increase the max number of hotlinks in ODB , defined as 
> 
> #define MAX_OPEN_RECORDS       256           /**< number of open DB records   */
> 
> I started running into an assertion in midas/src/odb.cxx
> 
> https://bitbucket.org/tmidas/midas/src/fa5457b5274a6b42c5ed8b6dea5e3cdd43de38fe/src/odb.cxx#lines-1525 :
> 
>   assert(sizeof(DATABASE_CLIENT) == 2112);
> 
> is it possible that the size of the DATABASE_CLIENT structure should be checked against 64+sizeof(OPEN_RECORD)*MAX_OPEN_RECORDS ?
> - 64 clearly can be expressed in a better maintainable form

Yes, this assert needs to be updated if you increase MAX_OPEN_RECORDS. See 
https://daq00.triumf.ca/MidasWiki/index.php/FAQ#Increasing_Number_of_Hot-links

> UPDATE: similar consideration holds for the size of the DATABLE_HEADER structure, which is also checked against a constant
> 
> https://bitbucket.org/tmidas/midas/src/fa5457b5274a6b42c5ed8b6dea5e3cdd43de38fe/src/odb.cxx#lines-1526

Yes DATABASE_HEADER can also be updated, but the associated assert()s need to be updated too.

FYI I have done both of these things and have attached patches.

Cheers,

Nick.

> when I attempted to increase the max number of hotlinks in ODB , defined as 
> #define MAX_OPEN_RECORDS       256           /**< number of open DB records   */
> assert(sizeof(DATABASE_CLIENT) == 2112);

Yes, it is intended to work like this. If you change MAX_OPEN_RECORDS (and some settings),
you break binary compatibility with standard MIDAS and the asserts inform you about it.

It is not a light step to take - you have to recompile all MIDAS clients, and if you miss
one and run it against your non-standard MIDAS, kaboom everything will go,
there is no safety net against this.

In the ALPHA experiment at CERN, for years we have been running with MAX_OPEN_RECORDS set to 2560,
and it works, you have to change both MAX_OPEN_RECORDS in midas.h and the expected values
in the assert() statements.

The new correct values you do not need to guess or compute yourself, the code to print
them is right there and it is easy to enable.

Replacing the numeric constants with computed values of course would completely defeat
the purpose of the tests - to catch the situation where by mistake or by ignorance
(or by miscompilation) sizes of critical data structures become different from those
normally expected.

K.O.

I see - thank you for the explanation!  

Indeed, updating MIDAS clients on each and every RPI etc in a running experiment may be a real challenge. 

Thinking forward - would it help if the ODB clients, upon initial connection but before doing anything else 
were reading the ODB parameters from the ODB itself, so the clients were "learning" about the ODB structure 
dynamically, at run time? Or that knowledge has to be static ?

-- thanks, regards, Pavel

I wonder why one needs more than 256 hotlinks at all. Please note that with the odbxx "watch" API, you can hotline a whole subdirectory, and get notified if ANY of the 
underlying values or subdirectories change. In principle, one could have one hotlink to "/" and see all changes in the ODB (although that does not make sense and might slow 
down ODB access a bit).

Try the odbxx_test.cpp example in MIDAS. In line 210 it puts a single hotlink to /Experiment. If you change anything under /Experiment, the program gets notified. By checking the 
path of the changed ODB entry, it can figure out which of the subways have been changed:

   // watch ODB key for any change with lambda function
   midas::odb ow("/Experiment");
   ow.watch([](midas::odb &o) {
      std::cout << "Value of key \"" + o.get_full_path() + "\" changed to " << o << std::endl;
   });


Maybe that would solve your problem without having to change the maximum number of hotlinks.

Stefan

> Indeed, updating MIDAS clients on each and every RPI etc in a running experiment may be a real challenge.

actually, only local clients must be rebuilt, remote clients connecting to the mserver do not care about ODB 
internal structure.

> Thinking forward - would it help if the ODB clients, upon initial connection but before doing anything else 
> were reading the ODB parameters from the ODB itself, so the clients were "learning" about the ODB structure 
> dynamically, at run time? Or that knowledge has to be static ?

unfortunately, the "open records" structure is allocated at compile-time inside the ODB header,
making any change to this would break binary compatibility.

I think it is possible to allocate "space for additional open records" in the ODB data area
and have the ODB open records code use it in addition to the compile-time allocated
space in the database header. This would also work for extending MAX_CLIENTS.

Of course in this approach, old midas clients would see only the clients and open records
in the database header, new midas clients would see the additional data.

It is not super hard to add this code...

K.O.

> I wonder why one needs more than 256 hotlinks at all.

I confirm that ALPHA is running with MAX_OPEN_RECORDS changed from 256 to 2048,
this is the only experiment I know of that had to increase any MIDAS ODB defaults.

The reason for this is mlogger, it opens an open record for each variable in each equipment.

This should be changed to 1 db_watch per equipment. We talked about it, but I guess we never did it.

I think this task just went almost to the top of my MIDAS to-do list.

K.O.

For the record, the ND280 (T2K near detector) MIDAS GSC was initially set up
with MAX_OPEN_RECORDS = 2560 and MAX_CLIENTS = 128.

In 2023 one fairly simple part of the detector was replaced with several 
other more complex systems (many more midas frontends, equipments, and
variables being logged) so we updated MAX_OPEN_RECORDS = 4096 and 
MAX_CLIENTS = 256.

Nick.

> I wonder why one needs more than 256 hotlinks at all. Please note that with the odbxx "watch" API, you can hotline a whole subdirectory, and get notified if ANY of the 
> underlying values or subdirectories change. In principle, one could have one hotlink to "/" and see all changes in the ODB (although that does not make sense and might slow 
> down ODB access a bit).

Thanks ! - I didn't know that. I did run into a number of hotlinks limit via mlogger which complained about not being able to create a hotlink 
to yet another event. Doubling the default value of MAX_OPEN_RECORDS solved the problem. 

I don't know the exact arithmetic defining the number of hotlinks in the system, but my today's case is a case of 
- 36 (linux servers) +18 (RPI) monitoring frontends managing one or several different equipment items each. 
- Each equipment item sends to ODB at least one monitoring event 
- in addition, each frontend created an individual hotlink for handling interactive commands 
- for MAX_OPEN_RECORDS=256, 4 equipment items per frontend easily make it into the dangerous zone. 

"Equipment items" also include the online processes running on the distributed computing farm processing the data .. 
(we are not using MIDAS event building capabilities)
 
> 
> Try the odbxx_test.cpp example in MIDAS. In line 210 it puts a single hotlink to /Experiment. If you change anything under /Experiment, the program gets notified. By checking the 
> path of the changed ODB entry, it can figure out which of the subways have been changed:
> 
>    // watch ODB key for any change with lambda function
>    midas::odb ow("/Experiment");
>    ow.watch([](midas::odb &o) {
>       std::cout << "Value of key \"" + o.get_full_path() + "\" changed to " << o << std::endl;
>    });
> 
> 
> Maybe that would solve your problem without having to change the maximum number of hotlinks.

I'll see how much mileage one can make here, but so far it looks that it is the number of various monitoring events 
handled by the mlogger which  drives the number of hotlinks 

-- thanks, regards, Pavel

> > I wonder why one needs more than 256 hotlinks at all.
> 
> I confirm that ALPHA is running with MAX_OPEN_RECORDS changed from 256 to 2048,
> this is the only experiment I know of that had to increase any MIDAS ODB defaults.
> 
> The reason for this is mlogger, it opens an open record for each variable in each equipment.
> 
> This should be changed to 1 db_watch per equipment. We talked about it, but I guess we never did it.
> 
> I think this task just went almost to the top of my MIDAS to-do list.

I definitely had many more than 256 variables successfully monitored with MAX_OPEN_RECORDS=256.
Is it possible that mlogger creates a hotlink per monitoring event, not per variable ? 
- I think, that would make more sense in almost any scenario... 

-- thanks, regards, Pavel

> > Indeed, updating MIDAS clients on each and every RPI etc in a running experiment may be a real challenge.
> 
> actually, only local clients must be rebuilt, remote clients connecting to the mserver do not care about ODB 
> internal structure.

thanks! I see - local clients do know about the memory mapping, remote ones - don't  

> unfortunately, the "open records" structure is allocated at compile-time inside the ODB header,
> making any change to this would break binary compatibility.

right, I guess, what I had in mind would require the very first fODB record to be a format descriptor, 
and that would be a breaking change... Anyway, the practical part of the problem is addressed, 
so I just add here a link which contains an answer to the original posting (I found it only after the fact):
 
https://daq00.triumf.ca/MidasWiki/index.php/FAQ#Increasing_Number_of_Hot-links

-- thanks again, regards, Pavel

Dear all,

I have problem in event buffer size.

When run MIDAS, I got error "total event size (1307072) larger than buffer size
(1048576)", so I guess that the EVENT_BUFFER_SIZE is small.

I change EVENT_BUFFER_SIZE in midas.h from 0x100000 to 0x200000. After compiling
and run MIDAS, I got other error "Shared memory segment with key 0x4d040761
already exists, please remove it manually: ipcrm -M 0x4d040761 size0x204a3c" in
system.C

I check the shmget() function in system.C and it is said that error come from
Shared memory segments larger than 16,773,120 bytes and create teraspace shared
memory segments

Anyone has this problem before?
Thanks for your help

M.T

Dear all,

I have problem in event buffer size.

When run MIDAS, I got error "total event size (1307072) larger than buffer size
(1048576)", so I guess that the EVENT_BUFFER_SIZE is small.

I change EVENT_BUFFER_SIZE in midas.h from 0x100000 to 0x200000. After compiling
and run MIDAS, I got other error "Shared memory segment with key 0x4d040761
already exists, please remove it manually: ipcrm -M 0x4d040761 size0x204a3c" in
system.C

I check the shmget() function in system.C and it is said that error come from
Shared memory segments larger than 16,773,120 bytes and create teraspace shared
memory segments

Anyone has this problem before?
Thanks for your help

M.T

> I have problem in event buffer size.
> 
> When run MIDAS, I got error "total event size (1307072) larger than buffer size
> (1048576)", so I guess that the EVENT_BUFFER_SIZE is small.
>

Correct. You have a choice of sending smaller events or increasing the buffer size.

Increasing the buffer size consumes computer memory, how much memory do you have on your machine?

> 
> I change EVENT_BUFFER_SIZE in midas.h from 0x100000 to 0x200000. After compiling
> and run MIDAS, I got other error "Shared memory segment with key 0x4d040761
> already exists, please remove it manually: ipcrm -M 0x4d040761 size0x204a3c" in
> system.C
> 

This is not normal. In recent versions of MIDAS (for the last few years)

a) buffer size is changed via ODB "/Experiment/buffer sizes", no need to edit midas.h
b) shared memory was switched from SYSV shared memory to POSIX shared memory, and you should not see any references to 
SYSV shared memory functions like "ipcrm", "shmget" and "segment key".

Are you using a very old version of MIDAS? Or maybe you have a MIDAS installation that still uses SYSV shared memory. Check 
the contents of .SHM_TYPE.TXT (in the same directory as .ODB.SHM), if would normally say "POSIXv2_SHM". If it says 
something else, it is best to convert to POSIX SHM. Simplest way is to stop everything, save odb to text file, delete 
.SHM_TYPE.TXT, restart odb with odbedit, reload from text file. Now check that .SHM_TYPE.TXT says "POSIXv2_SHM".

>
> I check the shmget() function in system.C and it is said that error come from
> Shared memory segments larger than 16,773,120 bytes and create teraspace shared
> memory segments
> 

What teraspace?!? You changed the size from 1 Mbyte to 2 Mbyte (0x200000), this is still below even the value you have above 
(16,773,120).

At the end, it is not clear what your problem is. After changing the shared memory size (via odb or via midas.h),
the midas *will* complain about the mismatch in size (existing vs expected) and will tell you how to fix it, (run "ipcrm").
After does this, is there still an error? Normally everything will just work. (you might also have to erase .SYSTEM.SHM,
midas will tell you to do so if it is needed).

So what is your final error? (After running ipcrm?)

K.O.

I am sorry for my late reply memory in my PC is 16 GB I check the contents of .SHM_TYPE.TXT and it is "POSIXv2_SHM". But there is no buffer sizes in "/Experiment" After run "ipcrm -M 0x4d040761 size0x204a3c", remove .SYSTEM.SHM and run MIDAS again, I still get error "Shared memory segment with key 0x4d040761 already exists, please remove it manually: ipcrm -M 0x4d040761 size0x204a3c" M.T

I am sorry for my late reply 

memory in my PC is 16 GB 

I check the contents of .SHM_TYPE.TXT and it is "POSIXv2_SHM". 
But there is no buffer sizes in "/Experiment" 

After run "ipcrm -M 0x4d040761 size0x204a3c", remove .SYSTEM.SHM and run MIDAS again, I still get error "Shared memory segment
with key 0x4d040761 already exists, please remove it manually: ipcrm -M 0x4d040761 size0x204a3c" M.T

> memory in my PC is 16 GB 

You can safely go to buffer size 100 Mbytes or more.

> I check the contents of .SHM_TYPE.TXT and it is "POSIXv2_SHM".

Good.

> But there is no buffer sizes in "/Experiment" 

This is strange. How old is your midas? What does it say on the "help" page in "Revision"?

> After run "ipcrm -M 0x4d040761 size0x204a3c"

This command is wrong. It probably gave you an error instead of removing the shared memory, that's why
nothing worked afterwards.

My copy of system.c reads this:
cm_msg(MERROR, "ss_shm_open", "Shared memory segment with key 0x%x already exists, please remove it manually: ipcrm -M 0x%x", key, key);

Note how there is no text "size0x..." in my copy? What does your copy say? Did somebody change it?

> remove .SYSTEM.SHM and run MIDAS again, I still get error "Shared memory segment
> with key 0x4d040761 already exists, please remove it manually: ipcrm -M 0x4d040761 size0x204a3c" M.T

Yes, that's because the ipcrm command is wrong and did not work,
it should read "ipcrm -M 0x4d040761" without the spurious "size..." text.

K.O.

> memory in my PC is 16 GB 

You can safely go to buffer size 100 Mbytes or more.

> I check the contents of .SHM_TYPE.TXT and it is "POSIXv2_SHM".

Good.

> But there is no buffer sizes in "/Experiment" 

This is strange. How old is your midas? What does it say on the "help" page in "Revision"?

> After run "ipcrm -M 0x4d040761 size0x204a3c"

This command is wrong. It probably gave you an error instead of removing the shared memory, that's why
nothing worked afterwards.

My copy of system.c reads this:
cm_msg(MERROR, "ss_shm_open", "Shared memory segment with key 0x%x already exists, please remove it manually: ipcrm -M 0x%x", key, 
key);

Note how there is no text "size0x..." in my copy? What does your copy say? Did somebody change it?

> remove .SYSTEM.SHM and run MIDAS again, I still get error "Shared memory segment
> with key 0x4d040761 already exists, please remove it manually: ipcrm -M 0x4d040761 size0x204a3c" M.T

Yes, that's because the ipcrm command is wrong and did not work,
it should read "ipcrm -M 0x4d040761" without the spurious "size..." text.

K.O.

> > memory in my PC is 16 GB 
> 
> You can safely go to buffer size 100 Mbytes or more.
> 
> > I check the contents of .SHM_TYPE.TXT and it is "POSIXv2_SHM".
> 
> Good.


No, wait, this is all wrong. If it says POSIX shared memory, how come it later
complains about SYSV shared memory and tells you to run SYSV shared memory
commands like ipcrm?!?


> > But there is no buffer sizes in "/Experiment" 


Now this kind of makes sense - you are probably running a strange mixture
of very old and recently new MIDAS. Probably you current version is so old
that it does not use .SHM_TYPE.TXT and can only do SYSV shared memory
and so old it does not have "/Experiment/buffer sizes".

But at some point you must have run a recent version of midas, or you would
not have the file .SHM_TYPE.TXT in your experiment directory.

I say:

a) run the correct ipcrm command (without the spurious "size..." text)
b) review your computer contents to identify all the versions of midas
   and to make sure you are using the midas you want to use (old or new,
   whatever), but not some wrong version by accident (incorrect PATH setting, etc)

As MIDAS developers, we usually recommend that you use the latest version of MIDAS,
certainly latest version is simpler to debug.

K.O.

The ROOTANA midas analyzer uncovered a problem with recursive use of ss_suspend().

When running in graphical mode, the ROOT graphics main event loop was calling 
ss_suspend(0) (no MSG_BM) which would sometimes call the user event handler (if a new 
event arrives into the midas event buffer). Because this loop was already running in the 
user event handler, there was a crash.

This is the calling sequence leading to the incorrect recursion: (from system.cxx comments 
to ss_suspend())

analyzer ->
     -> cm_yield() in the main loop
     -> ss_suspend(0)
     -> MSG_BM message arrives in the UDP socket
     -> ss_suspend_process_ipc(0)
     -> cm_dispatch_ipc()
     -> bm_push_event()
     -> bm_push_buffer()
     -> bm_read_buffer()
     -> bm_dispatch_event()
     -> user event handler
     -> user event handler ROOT graphics main loop needs to sleep
     -> ss_suspend(0) <--- should be ss_suspend(MSG_BM)!!!     
     -> MSG_BM message arrives in the UDP socket
     -> ss_suspend_process_ipc(0) <- should be ss_suspend_process_ipc(MSG_BM)!!!
     -> cm_dispatch_ipc() <- without MSG_BM, calling cm_dispatch_ipc() again
     -> bm_push_event()
     -> bm_push_buffer()
     -> bm_read_buffer()
     -> bm_dispatch_event()
     -> user event handler <---- called recursively, very bad!

The proper fix for this is to always call ss_suspend(MSG_BM) from the user event handler 
and ss_suspend(MSG_ODB) from the user db_watch handler.

In this second case, ss_suspend(MSG_OBM) will lose/ignore/discard db_watch notifications, 
if you do not want that, call ss_suspend(0) and be ready for recursive calls to your 
db_watch handler. (this can happen in a frontend program that acts on changes in ODB and 
where some of these actions may require sleeping via ss_suspend()).

ss_suspend(MSG_BM) will discard MSG_BM messages, which is not a problem as 
bm_wait_for_events() and cm_yield() will immediately poll the event buffer and there will be 
no delay in receiving new events.

Until commit 99d6e90 there were problems in ss_suspend(MSG_BM) and recursive calls to 
the user event handler were still possible. It is now fixed in this and the previous commits.

K.O.

Hi,

Attempting to copy a subtree to a new location in the ODB using odbedit with "copy <src> <dest>" is 
occasionally not copying the entire <src> subtree.

I am experiencing this issue consistently when trying to copy subtrees from the "/Equipment" ODB tree to 
a new location.  The first 2-3 variables/directories of the <src> subtree will be copied to <dest> but the 
full subtree will not be copied over.