Bug#312810: Apache2 consumes 100% CPU after several LDAP authenticated requests

2005-06-23 Thread Geert Nijpels
We didn't see the problem anymore after we disabled HT and upgraded to
2.0.54-4 (we are running this for about a week). I'm OK with closing
this bug, unless you want to try and debug the HT problem.


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#312810: Apache2 consumes 100% CPU after several LDAP authenticated requests

2005-06-11 Thread Geert Nijpels
Adam Conrad wrote:

Geert Nijpels wrote:
  

It looks like this problem is related to bug #307567. Apache starts
consuming 100% CPU after some requests. Restarting Apache lets it serve
pages for a few minutes again.

Downgrading to apache* 2.0.54-2 fixed the problem. We use
apache2-mpm-prefork on an up-to-date sarge installation.



Well, that's just bizarre, since 2.0.54-2 is the version with the new
util_ldap code from upstream, while 2.0.54-4 is completely reverted to the
tried and tested code from 2.0.53 and previous.  Everyone else (so far)
has found 2.0.54-4 to be much more stable and useable.

Are you pushing a LOT of LDAP requests through apache2?  If so, I could
see how the 2.0.53 (2.0.54-4) code might eventually lead to a bunch of
hung connections (though, it's always been like this), but you'd have to
be pushing a LOT of LDAP requests through, or have a serious CPU/RAM
shortage for it to spike your system really badly.
  

We are not pushing a lot of LDAP requests, something like 10/minute, and
CPU / RAM should be enough.
It's a dual xeon 2.00GHz with hyperthreading enabled, memory usage:

matrix:/home/geert# cat /proc/meminfo
MemTotal:  1035908 kB
MemFree: 99436 kB
Buffers:164168 kB
Cached: 465308 kB
SwapCached:  0 kB
Active: 405092 kB
Inactive:   388248 kB
HighTotal:  130496 kB
HighFree:  140 kB
LowTotal:   905412 kB
LowFree: 99296 kB
SwapTotal: 1951888 kB
SwapFree:  1951876 kB
Dirty: 960 kB
Writeback:   0 kB
Mapped: 181368 kB
Slab:   131984 kB
CommitLimit:   2469840 kB
Committed_AS:   608536 kB
PageTables:   1924 kB
VmallocTotal:   114680 kB
VmallocUsed: 21152 kB
VmallocChunk:93364 kB

I built the same setup on a second server on which I can not reproduce
it despite doing 1000's of requests. This server does not do
hyperthreading and has a single CPU. So I guess the next thing I will
try is disabling hyperthreading and go back to 2.0.54-4.

We are running:
2.6.11.11 #1 SMP

Geert


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#312810: Apache2 consumes 100% CPU after several LDAP authenticated requests

2005-06-10 Thread Geert Nijpels
Package: apache2-utils
Version: 2.0.54-4

It looks like this problem is related to bug #307567. Apache starts
consuming 100% CPU after some requests. Restarting Apache lets it serve
pages for a few minutes again.

Downgrading to apache* 2.0.54-2 fixed the problem. We use
apache2-mpm-prefork on an up-to-date sarge installation.

Backtrace is below.

gdb) backtrace
#0  0xb7e5d2e3 in apr_date_parse_rfc () from /usr/lib/libaprutil-0.so.0
#1  0xb7e5d4e8 in apr_date_parse_rfc () from /usr/lib/libaprutil-0.so.0
#2  0xb7e5db0c in apr_rmm_calloc () from /usr/lib/libaprutil-0.so.0
#3  0xb7a79a72 in util_ald_create_cache ()
   from /usr/lib/apache2/modules/mod_ldap.so
#4  0xb7a79968 in util_ald_create_caches ()
   from /usr/lib/apache2/modules/mod_ldap.so
#5  0xb7a777ec in util_ldap_cache_checkuserid ()
   from /usr/lib/apache2/modules/mod_ldap.so
#6  0xb7a6ebd5 in mod_auth_ldap_check_user_id ()
   from /usr/lib/apache2/modules/mod_auth_ldap.so
#7  0x08090035 in ap_run_check_user_id ()
#8  0x08090a2c in ap_process_request_internal ()
#9  0x08069c88 in ap_process_request ()
#10 0x0806512d in _start ()
#11 0x08a9fbf0 in ?? ()
#12 0x0004 in ?? ()
#13 0x08a9fbf0 in ?? ()
#14 0xb7a38d80 in modperl_process_connection_handler ()
   from /usr/lib/apache2/modules/mod_perl.so
#15 0x080835f5 in ap_run_process_connection ()
#16 0x080769a4 in ap_graceful_stop_signalled ()
#17 0x08076af7 in ap_graceful_stop_signalled ()
#18 0x08076d89 in ap_graceful_stop_signalled ()
#19 0x08077475 in ap_mpm_run ()
#20 0x0807dabd in main ()
(gdb) 




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]



Bug#312810: Apache2 consumes 100% CPU after several LDAP authenticated requests

2005-06-10 Thread Adam Conrad
Geert Nijpels wrote:

 It looks like this problem is related to bug #307567. Apache starts
 consuming 100% CPU after some requests. Restarting Apache lets it serve
 pages for a few minutes again.

 Downgrading to apache* 2.0.54-2 fixed the problem. We use
 apache2-mpm-prefork on an up-to-date sarge installation.

Well, that's just bizarre, since 2.0.54-2 is the version with the new
util_ldap code from upstream, while 2.0.54-4 is completely reverted to the
tried and tested code from 2.0.53 and previous.  Everyone else (so far)
has found 2.0.54-4 to be much more stable and useable.

Are you pushing a LOT of LDAP requests through apache2?  If so, I could
see how the 2.0.53 (2.0.54-4) code might eventually lead to a bunch of
hung connections (though, it's always been like this), but you'd have to
be pushing a LOT of LDAP requests through, or have a serious CPU/RAM
shortage for it to spike your system really badly.

At any rate, given everyone else's track record so far with these
versions, I think it's safe to say that the version in Sarge will stay as
it is, however in Sid we will be getting apache2 2.0.55 soon, which has
the newer code (with LDAP connection timeouts), but it's actually meant to
work now, rather than randomly segfaulting and spiking the CPU.

... Adam




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of unsubscribe. Trouble? Contact [EMAIL PROTECTED]