[TYPO3-english] backend hangs randomly

[Bjoern Pedersen]

Am 22.02.2011 11:40, schrieb Nicola Cerchiari:
> Hi to all,
> First time here suggested by friends of the italian mialing list; i'm
> encountering a very annoying problem of which I can find no solutions.
> 
> The problem is that randomly during day calling any backend action hang the
> browser on "waiting for response..." and the only solution left is to stop
> the call, wait several minutes and then everything is back to normal until
> it happens again.
> For example the same action works for 2 hours and then all of a sudden
> hangs. Even accessing to backend login page sometimes hangs.
> I've tested the same website and core on my test machines and I never
> encounter any problem.
> I give you some data in other to better understand the situation:
> - php 5.2.x with all libraries needed
> - mysql 5.1.55 in replication master/slave (just updated)
> - 2 balanced webserver apache/redhat (on a vmware virtualized farm)
> - filesystem of webserver ron a shared NAS
> - version of typo3 on which the problem is encountered: 4.4.6 and 4.5.0
> 
> The matter is that when backend works, works pretty fast, but the nall of a
> sudden with apparently no reason it hangs.
> 
> System admnistrators of the infrastructure suggested at first that there was
> a "strangly" large amount of calls in TIME_WAIT state, then they suggested
> maybe the mysql server is not tuned sufficiently to stand the increasing
> requests (consider that this infrastructure receive >50.000 user visit a
> day) and they will try a fine tuning in the next day.
> 
> What i'm looking for is to understand if the problem is due to Typo3 (i
> don't think so) or there is some problem in our infrastructure.
> 
> Any ideas?
> 
> 
> Thank you in advance,
> Nicola

Possible causes: clearing cache: lots of rather slow delete queries on
db -> there is a extension to use truncate table for clearing the cache.

stalling locking ( there are some race conditions left in the locking
code, check the  bugtracker)

DB-backup not excluding the cache tables: huge amount of data with table
locking, so lots of db request waiting foe the lock to become released.

Björn