We just encountered an 'interesting' problem that I thought I'd
mention. We discovered that our memory/swap utilization was
slowly growing over a period of several weeks. Checking the
size of all our procs [with ps -l], we were amazed to see that the backends
had grown from some 1900 blocks, to over 15,000 in some cases [YES, almost
an order of magnitude]. Naturally, performance decreased, and some
machines even crashed.
Checking with Informix, I was notified that using SIMPLE SELECTS
will cause this occur. [I.E. - this is a KNOWN bug].
I'm waiting for a fix, since I have too much code to change.
If I get more info, I'll post.
Standard disclaimers,
Ihor Kinal
att!cbnewsh!ijk [netnews address]
att!trumpet!ijk [real mail address]
P.S. We're seeing some growth in our front-ends, as well.
Of course, here it's more difficult to determine if this is
related to any ESQL calls, or the direct mis-use of malloc.