I'm afraid I'm missing something... Can you better define the slowness? Do
the sub-second queries take longer now? Or do you believe it's the
connection establishment that takes longer?
Make sure you follow John's suggestions... A few more questions:
- Are you sure ther is no sysdbopen() procedure being run?
- How many connects per second do you have? (audit facility can tell you
- Why do you have 40 poll threads?
- Do you already have NUMFDSERVERS?
- Do you see many sessions with flag "S" in onstat -u? If yes, do you see
nsf lock very "hot"? (onstat -g wmx or onstat -g lmx)
- How is your ready queue? (onstat -g rea -r 1)
- How is the load on the machine?
- If the load is too high, are the oninits causing it? And if yes, which
oninits? (CPUVPs or the poll threads?
- Can you check the quick queries query plan? Do they look ok?
Post by Laurie Gustin
We have recently migrated from IDS 11.5 to IDS11.70 FC5 instance running
on Linux. We have one application that makes many separate connections to
the database. They connect, run small queries and disconnect. The queries
usually take less than one second, so these connections come and go very
quickly and there are alot of them.
We are experiencing some slowness on the 11.7 server that we did not see
on 11.5, so I'm looking for help getting this new server tuned up. The
only think I know to look at is NETTYPE params
11.5 - NETTYPE soctcp, 50, 100, NET - when I start this engine I get
this error: Warning: Number of poll threads for nettype soctcp has already
reached the maximum of 50
11.7 - NETTYPE soctcp, 40, 100, NET
Thanks in advance!
Informix-list mailing list
My email works... but I don't check it frequently...