Cloverleaf 5.8 recovery and error database size max

[Steve Herber]

Other clovertech posts, and my own opinion, were that the recovery and error databases were limited to 2GB.  Testing on AIX 6.1 was stopped after the size reached 6.4GB:

ls -l rlogM2k.dat                                    

-rw-rw-r–    1 hci      staff    6446481408 Mar 16 16:08 rlogM2k.dat

with about 2,110,000 test ADT records in the database.

What are the limits for cloverleaf 5.8?

What are the limits for cloverleaf 6.1?

Thanks!

Steve Herber

Steve Herber
University of Washington

Topic

[Russ Ross]

I think a large part of file size limits are controlled by the OS settings and not that much by the application, so I’m curious what your /etc/security/limits file has for fsize setting for the hci user or the default.

In AIX I beleive it is the fsize setting in the file called /etc/secuirty/limits.

I cheked ours and see the setting is 2,097,151 (512 byte blocks) which comes to 1GB.

That coinsides with what I saw the other day when I couldn’t FTP a 1.3 Gb file until I reduced it below 1GB in size.

I was caught off guard and thought for some reason my limit was 2GB maybe because the fsize settings is in 512 byte blocks, so I have to divide it by 2 to convert it to megabytes.

Honestly a 6.4 GB files size would cause me to raise other questions about necessity, like is the site too large, or why so many queued messages, that by the way will aslo use up precious RAM since a copy resides in memory until delivered.

on AIX once you use up all your RAM and message start thrashing in paging space, Cloverleaf gets so slow it is next to unuseable.

We only had 4 GB of RAM and have recently increased it to 16 GB we need for our Epic project that has us using 10 GB of RAM as of today.

This command is what I use on AIX to check if I’m using any paging space because if I used that up then my system crashes hard and fails over:

lsps -a

and here is a script ( show_memory_used.ksh ) I wrote to show me how much my memory demands are, so I know how close I’m getting to running out before I start using paging space:

Code:

#!/usr/bin/ksh

echo `ps -elfZ | awk ‘{TOTAL+=$10} END {print (TOTAL/1024/1024)}’` GB

Also be aware when we were on cloverleaf 5.6.? we had overlooked a que that was 627K messages deep without adverse impact.

After upgrading to cloverleaf 6.0 it is smarter to not put up with such abused and we saw the engine stop taking messages on the inbound thread for that site once an outbound threads hit queue depth of 80K due to an interface that was too slow (1 message per 5 seconds).

We had to put these bad interfaces in their own sites to create an additional layer to protect the good interfaces.

I don’t think later version of cloverleaf are as tolerant of abuse anywhere close to what you are describing as it will try and go into a self defense mode quicker than older versions of cloverleaf.

Please note this is a deduced reasoning from my personal observation while going to cloverleaf 6.0.

This is sort of similar to how the engine will toggle off SMAT if sensing disk space is about to run out but now it might be doing something similar for message queue depths regardless of how large of a file you can right to disk.

Russ Ross
RussRoss318@gmail.com

[Jon Blanchard]

Russ:

How much hard disk space did you allocate on the engine server to accommodate the Epic installation?

Thanks.

Migration Consultant
Coffee Regional Medical Center
Eastern Time Zone

[Author]

Replies