[Info-ingres] A knotty problem with a checkpoint procedure.

[user1region1 at gmail.com]

On Tuesday, June 20, 2017 at 11:14:03 AM UTC+1, Martin Bowes wrote:
> What do you have connect_limit set to?
> 
> I ask because the only thing I can find regarding the error E_CL0654 mentions this as a possibility. Why it would affect just Solaris work Scheduler is a mystery.


We have:

ii.sysa.dbms.*.connect_limit:           256

5 iidbms processes running.


This was reduced previously from connect_limit=512, again for five servers.


After looking at the truss output (albeit that it worked) I added this to the test script just now:

ulimit -Hn
ulimit -Sn
ulimit -Hn 65536
ulimit -Sn 65536


The result was that we saw:

+ ulimit -Hn
65536
+ ulimit -Sn
256
+ ulimit -Hn 65536
+ ulimit -Sn 65536


Plus, the flipping "infodb" worked this time when NOT run through truss.


So that seems to be the issue. There is something screwy here. Quite whose fault it is I don't know, but it would be nice to get it sorted out. I don't see why customers (of Actian, Oracle or Automic) should have to muck about like this, but I don't want them all playing musical chairs!