malloc issue?
Fairlight
fairlite at fairlite.com
Mon May 20 07:48:15 PDT 2019
*** Error in `rclerk': malloc(): memory corruption (fast): 0x08832cb8 ***
We're hitting this when memory is -not- exhausted. There's plenty of disk
cache to cannibalise, as well as SLAB memory to cannibalise. Usually this
happens when we get down into the 350-550MB free range (not dropping below
350 on this particular server). Today, it started kicking these errors out
with 2.5GB of RAM free, which should have been fairly well out of memory
contention range.
filePro 5.0.14
CentOS 7 with kernel 5.0.7 (kernel doesn't even matter; it's been upgraded
several times since we started observing this).
What exactly is "fast" malloc()? TTBOMK, there is only one glibc-based
malloc(), full stop. Is this a custom one coded into fP which somehow
bypasses the usual system memory reclamation which would normally be
triggered by a call to the real malloc()? That's my working hypothesis,
having seen a bunch of custom malloc() implementations in various software
over the years. If so, it's a horrible practise.
What's the actual solution to the issue, given that the system actualy
-has- free RAM to use, on the order of hundreds of times more than the
32-bit image could possibly address at once? As it is, the only way I've
found to rectify the issue is to drop caches on the ground to free up more
RAM than should be necessary.
Thanks,
mark->
--
Fairlight Consulting
http://www.fairlite.com
fairlite at fairlite.com
(502) 509-3840
More information about the Filepro-list
mailing list