Mailing List Archive: strange Locking behavior on OS-X Server

strange Locking behavior on OS-X Server

Mar 21, 2007, 9:26 AM

Post #1 of 3 (2542 views)

Hello,

on OS-X Server, Version 10.4,
Apache/1.3.33 (Darwin) Embperl/2.2.0 DAV/1.0.3 mod_ssl/2.8.24 OpenSSL/
0.9.7l mod_perl/1.29

a strange behaviour occurs when using "File" as storaage and either
"Semaphore" or "File" as locking, under certain circumstances mostly
under a minimum load with at least 3-5 concurrent requests, a request
may take several seconds (even a simple "hello" page) or lock
forever. Killing some of the waiting processes or restarting the
server helps for a while until the (for me unknown) race-condition
occurs again...

After setting the lock-method to "Null" this situation did no more
occur again but now a potential session-inconstency could occur which
does not satisfy me.

Did somone else also have such an odd behavior?

Regards,

Wolfgang Kinkeldei

--

' /\_/\ ' .print[split??,"".(($/=q|Cms)+-03467:;<=|)=~tr!C-z -B! -z!)x
'( o.o )' .$/]->[hex]foreach split qr<>,qq+1ecd039ad65b025b8063475b+||
' > ^ < ' .q

RE: strange Locking behavior on OS-X Server [ In reply to ]

richter at ecos

Apr 2, 2007, 7:08 AM

Post #2 of 3 (2342 views)

Permalink

Hi,

>
> on OS-X Server, Version 10.4,
> Apache/1.3.33 (Darwin) Embperl/2.2.0 DAV/1.0.3 mod_ssl/2.8.24
> OpenSSL/ 0.9.7l mod_perl/1.29
>
> a strange behaviour occurs when using "File" as storaage and
> either "Semaphore" or "File" as locking, under certain
> circumstances mostly under a minimum load with at least 3-5
> concurrent requests, a request may take several seconds (even
> a simple "hello" page) or lock forever. Killing some of the
> waiting processes or restarting the server helps for a while
> until the (for me unknown) race-condition occurs again...
>
> After setting the lock-method to "Null" this situation did no
> more occur again but now a potential session-inconstency
> could occur which does not satisfy me.
>
> Did somone else also have such an odd behavior?
>

I can't help you with this issue, but unless you use frame, where the
same session key might be used in parallel, it#s normaly safe to use the
Null locker.

Otherwise you might try to use mysql for data storeage and locking, as
far as I know, this works without problems.

Gerald

P.S. Do you use %mdat or only %udat for sessions?

---------------------------------------------------------------------
To unsubscribe, e-mail: embperl-unsubscribe@perl.apache.org
For additional commands, e-mail: embperl-help@perl.apache.org

Re: strange Locking behavior on OS-X Server [ In reply to ]

wolfgang at kinkeldei

Apr 5, 2007, 7:55 AM

Post #3 of 3 (2333 views)

Permalink

Am 02.04.2007 um 16:08 schrieb Gerald Richter - ECOS GmbH:

> Hi,
>
>>
>> on OS-X Server, Version 10.4,
>> Apache/1.3.33 (Darwin) Embperl/2.2.0 DAV/1.0.3 mod_ssl/2.8.24
>> OpenSSL/ 0.9.7l mod_perl/1.29
>>
>> a strange behaviour occurs when using "File" as storaage and
>> either "Semaphore" or "File" as locking, under certain
>> circumstances mostly under a minimum load with at least 3-5
>> concurrent requests, a request may take several seconds (even
>> a simple "hello" page) or lock forever. Killing some of the
>> waiting processes or restarting the server helps for a while
>> until the (for me unknown) race-condition occurs again...
>>
>> After setting the lock-method to "Null" this situation did no
>> more occur again but now a potential session-inconstency
>> could occur which does not satisfy me.
>>
>> Did somone else also have such an odd behavior?
>>
>
> I can't help you with this issue, but unless you use frame, where the
> same session key might be used in parallel, it#s normaly safe to
> use the
> Null locker.
>
> Otherwise you might try to use mysql for data storeage and locking, as
> far as I know, this works without problems.

In the meantime, we found that under OS-X Server (PPC-Edition) very
strange things are happening with the whole session management
resulting in session-mixing. This mix occurs using "File" or "MySQL"
as a session storage and is independent on the kind of Locking or use
no Locking at all. On the Intel Edition I could not verify such a
behavior. As soon as there is more time we will try to track that
down. All Perl-modules we use do successfully complete their tests
without any complains.

> P.S. Do you use %mdat or only %udat for sessions?

I only use %udat for sessions.

Grüße,

Wolfgang

--

' /\_/\ ' .print[split??,"".(($/=q|Cms)+-03467:;<=|)=~tr!C-z -B! -z!)x
'( o.o )' .$/]->[hex]foreach split qr<>,qq+1ecd039ad65b025b8063475b+||
' > ^ < ' .q

Mailing List Archive

Attached Files:

Attached Files: