Hi Tom, We can run some tests. I have a VMware "clone" of the server that I can use for testing. Its basically a mondoarchive backup/restore, so config should be 100%, and I can test most things in it. Hardwares obviously different, and I cant emulate 12Gb of ram, but I can at least get 1Gb on it again. I used the yum RPM version that Redhat supplied. Next week I'll try some other things, like putting the latest (non-redhat) RPM on there, and maybe read up a little on what your asking - as right now I've no clue. I'll also see if the MaxMemFree helps out at 1Mb. Adrian -----Original Message----- From: Tom Evans [mailto:tevans.uk@xxxxxxxxxxxxxx] Sent: 03 April 2009 16:21 To: users@xxxxxxxxxxxxxxxx Subject: RE: Apache memory hog On Fri, 2009-04-03 at 14:24 +0100, Adrian Marsh wrote: > Hi Tom, > > Well, at the moment the servers running very low on available memory: > > Mem: 12307068k total, 12121380k used, 185688k free, 286808k buffers > Swap: 2031608k total, 0k used, 2031608k free, 11081692k cached > > Hence my original question #3, on where the memorys going after httpd is > stopped/started (via a normal init.d stop/start). > > Yet none of that 12gb seems to be allocated anywhere. The processes > listed below are sorted by max usage. Since my last email (10mins?) > httpd has grown to: > > 10002 apache 15 0 297m 35m 5024 S 0.0 0.3 0:02.17 httpd > 10007 apache 15 0 296m 33m 4864 S 0.0 0.3 0:01.11 httpd > 10003 apache 16 0 295m 32m 5164 S 0.0 0.3 0:00.60 httpd > 10005 apache 15 0 295m 32m 4952 S 0.0 0.3 0:00.99 httpd > 10006 apache 15 0 295m 32m 5012 S 0.0 0.3 0:01.02 httpd > 10000 apache 15 0 294m 31m 5256 S 0.0 0.3 0:01.42 httpd > 4226 mysql 15 0 257m 31m 4264 S 0.0 0.3 9:31.30 mysqld > 10001 apache 15 0 290m 27m 5228 S 0.0 0.2 0:00.48 httpd > 10004 apache 15 0 289m 27m 5144 S 0.0 0.2 0:00.25 httpd > > To get that memory back I have to reboot the server. > > The yum update did change a number of core packages yes. I'd tested it > for a few weeks in a VMWARE virtual machine, to check functionality, but > hadn't realized that it might even cause the issues we see now with > regards to memory. > > Is there anyway to dig into the memory usage within those httpd > processes? > > Hmmm, that does look like a memory leak indeed. As this is production, I guess we can't run too many tests? It would be useful to rebuild apache with debug symbols, see if you can still reproduce the leak. If you can, perhaps run it under valgrind in single process mode (-X flag to apache). That would show up where any memory leaks are. Cheers Tom --------------------------------------------------------------------- The official User-To-User support forum of the Apache HTTP Server Project. See <URL:http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx " from the digest: users-digest-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx --------------------------------------------------------------------- The official User-To-User support forum of the Apache HTTP Server Project. See <URL:http://httpd.apache.org/userslist.html> for more info. To unsubscribe, e-mail: users-unsubscribe@xxxxxxxxxxxxxxxx " from the digest: users-digest-unsubscribe@xxxxxxxxxxxxxxxx For additional commands, e-mail: users-help@xxxxxxxxxxxxxxxx
|