[imp] runaway HTTPD/Memory problems

[Matthew Hilty]

Good Morning,
    The problem WAS definitely a result of misconfiguration in which the folders
variable was empty on servers.php, but the namespace was 'INBOX.'   I am shocked
(but not surprised) that I made such a ridiculous error, but I'm also a little
surprised that it was working at all without 'folders' being set corrctly.  I
remember not setting it correctly once in IMP 2.2 and the CPU went crazy!  The
good thing that came out of it, I guess, is a cron job that traps, traces, and
kills runaway process,  Fixing IMP so that it works well again is a plus too!

thanks,

Matthew

Quoting Eric Rostetter <eric.rostetter at physics.utexas.edu>:

> Quoting Matthew Hilty <mhilty at artic.edu>:
> 
> > Hi all,
> > 	I've just recently provided access to IMP 3.1 to a ~3000 user
> > community which had been using 2.2.8 for about a year without problems.
> > During tests, the problem I'm about to describe did not exhibit itself,
> > but under the load of our production environment, it turns out that we're
> > experiencing occasional runaway httpd processes that consume up to 500+Mb
> > of RAM, and which force the server to swap dramatically. 
> 
> I've only seen run-away processes when using add-on utils like wvhtml,
> aspell, etc.  Any chance these might be involved?
> 
> If not, then judging by your trace below, I would guess that your imap
> folder settings might be off, and it is, perhaps, building a folder list
> consisting of all or most of the files on your system???  This is, I 
> admit, a wild guess...
> 
> > lstat64("/var/www/html/horde/imp/templates/folders",
> > {st_mode=S_IFDIR|0750, st_size=4096, ...}) = 0
> > lstat64("/var/www/html/horde/imp/templates/folders/row.inc",
> > {st_mode=S_IFREG|0640, st_size=461, ...}) = 0
> 
> Note it stops in folders/row.inc, hence my wild guess at a folder issue...
> 
> > ioctl(10, TCGETS, 0xbfff5940)           = -1 ENOTTY (Inappropriate ioctl
> > for device)
> 
> Could be it hit a special file or something here...
> 
> > fstat64(10, {st_mode=S_IFREG|0640, st_size=461, ...}) = 0
> > mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0)
> > = 0x44acb000
> > read(10, "<tr class=\"<?= $navclass ?>\">\n  "..., 8192) = 461
> 
> That is part of .../folders/row.inc
> 
> > read(10, "", 4096)                      = 0
> > read(10, "", 8192)                      = 0
> > ioctl(10, TCGETS, 0xbfff4910)           = -1 ENOTTY (Inappropriate ioctl
> > for device)
> > close(10)                               = 0
> > munmap(0x44acb000, 4096)                = 0
> 
> Again, just questions and wild guesses, no actual answers....
> 
> -- 
> Eric Rostetter
> The Department of Physics
> The University of Texas at Austin
> 
> Why get even? Get odd!
> 
> -- 
> IMP mailing list
> Frequently Asked Questions: http://horde.org/faq/
> To unsubscribe, mail: imp-unsubscribe at lists.horde.org
> 


-- 
Matthew Hilty
Network Administrator
The Art Institute of Chicago