drouting - dr_reload (performance)

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

drouting - dr_reload (performance)

Max Mühlbronner
Hello,



Opensips seems to not route my requests while reloading the drouting
rules from Database. Probably the DB operations are blocking the
remaining operations?

Any idea if this is normal behavior, or misconfiguration on my side?
Does anyone know a solution for reloading while still receiving traffic,
so if the rules become quite big it will not take the service down while
reloading via mi (fifo dr_reload / webinterface..)?


Best Regards

Max M.


_______________________________________________
Users mailing list
[hidden email]
http://lists.opensips.org/cgi-bin/mailman/listinfo/users
Reply | Threaded
Open this post in threaded view
|

Re: drouting - dr_reload (performance)

Bogdan-Andrei Iancu-2
Hi Max,

the reload is done in a separate process and only when done, the routing
data (in mem) is swapped, so while loading new data, opensips will still
use the old one - it should not block and wait for reload.

Are you sure is there any blocking  ? What version are you using ?

Regards,
Bogdan



On 06/16/2011 03:48 PM, Max Mühlbronner wrote:

> Hello,
>
>
>
> Opensips seems to not route my requests while reloading the drouting
> rules from Database. Probably the DB operations are blocking the
> remaining operations?
>
> Any idea if this is normal behavior, or misconfiguration on my side?
> Does anyone know a solution for reloading while still receiving
> traffic, so if the rules become quite big it will not take the service
> down while reloading via mi (fifo dr_reload / webinterface..)?
>
>
> Best Regards
>
> Max M.
>
>

--
Bogdan-Andrei Iancu
OpenSIPS solutions and "know-how"


_______________________________________________
Users mailing list
[hidden email]
http://lists.opensips.org/cgi-bin/mailman/listinfo/users
Reply | Threaded
Open this post in threaded view
|

Re: drouting - dr_reload (performance)

Max Mühlbronner
Hi,

yes, i tried several times. I should have mentioned the dr_rules has
quite a few (about 100k entries). I did notice it on a production server
running 1.6.2 but also was able to replicate the same behavior when
testing with the same Database/dr_rules on 1.6.4-tls.

I will try to take a deeper look into the issue, will try to send some
traces / logs if there is something which looks like it could be related.


BR

Max M.


Am 24.06.2011 16:40, schrieb Bogdan-Andrei Iancu:

> Hi Max,
>
> the reload is done in a separate process and only when done, the
> routing data (in mem) is swapped, so while loading new data, opensips
> will still use the old one - it should not block and wait for reload.
>
> Are you sure is there any blocking  ? What version are you using ?
>
> Regards,
> Bogdan
>
>
>
> On 06/16/2011 03:48 PM, Max Mühlbronner wrote:
>> Hello,
>>
>>
>>
>> Opensips seems to not route my requests while reloading the drouting
>> rules from Database. Probably the DB operations are blocking the
>> remaining operations?
>>
>> Any idea if this is normal behavior, or misconfiguration on my side?
>> Does anyone know a solution for reloading while still receiving
>> traffic, so if the rules become quite big it will not take the
>> service down while reloading via mi (fifo dr_reload / webinterface..)?
>>
>>
>> Best Regards
>>
>> Max M.
>>
>>
>


_______________________________________________
Users mailing list
[hidden email]
http://lists.opensips.org/cgi-bin/mailman/listinfo/users
Reply | Threaded
Open this post in threaded view
|

Re: drouting - dr_reload (performance)

Bogdan-Andrei Iancu-2
Hi Max,

100K rule is nothing for the dr module (disregarding the how fast the DB
is or not).

Maybe the so called locking is about overall load or DB load - using
some xlogs and benchmarking module, you can try to see if there is an
actual blocking in the do_routing() function.

Regards,
Bogdan

On 06/24/2011 06:13 PM, Max Mühlbronner wrote:

> Hi,
>
> yes, i tried several times. I should have mentioned the dr_rules has
> quite a few (about 100k entries). I did notice it on a production
> server running 1.6.2 but also was able to replicate the same behavior
> when testing with the same Database/dr_rules on 1.6.4-tls.
>
> I will try to take a deeper look into the issue, will try to send some
> traces / logs if there is something which looks like it could be related.
>
>
> BR
>
> Max M.
>
>
> Am 24.06.2011 16:40, schrieb Bogdan-Andrei Iancu:
>> Hi Max,
>>
>> the reload is done in a separate process and only when done, the
>> routing data (in mem) is swapped, so while loading new data, opensips
>> will still use the old one - it should not block and wait for reload.
>>
>> Are you sure is there any blocking  ? What version are you using ?
>>
>> Regards,
>> Bogdan
>>
>>
>>
>> On 06/16/2011 03:48 PM, Max Mühlbronner wrote:
>>> Hello,
>>>
>>>
>>>
>>> Opensips seems to not route my requests while reloading the drouting
>>> rules from Database. Probably the DB operations are blocking the
>>> remaining operations?
>>>
>>> Any idea if this is normal behavior, or misconfiguration on my side?
>>> Does anyone know a solution for reloading while still receiving
>>> traffic, so if the rules become quite big it will not take the
>>> service down while reloading via mi (fifo dr_reload / webinterface..)?
>>>
>>>
>>> Best Regards
>>>
>>> Max M.
>>>

--
Bogdan-Andrei Iancu
OpenSIPS solutions and "know-how"


_______________________________________________
Users mailing list
[hidden email]
http://lists.opensips.org/cgi-bin/mailman/listinfo/users
Reply | Threaded
Open this post in threaded view
|

Re: drouting - dr_reload (performance)

Max Mühlbronner
Hi,


I just found out what caused these issues, simple solution, thanks for
your hints!

If debug level is set to 0 it works just fine  + reload is really fast
(wether set via fifo debug or in config). At level3 you can see lots of
debugging messages when reloading, also the load goes up and the
blocking occurs.

At debug level 0 reload goes really fast and no blocking at all. So it
really seems to be related to the xlogs/debugging.

Everything is working fine for me now (at debug0). Thanks again.

BR

Max M.

Am 25.06.2011 13:54, schrieb Bogdan-Andrei Iancu:

> Hi Max,
>
> 100K rule is nothing for the dr module (disregarding the how fast the
> DB is or not).
>
> Maybe the so called locking is about overall load or DB load - using
> some xlogs and benchmarking module, you can try to see if there is an
> actual blocking in the do_routing() function.
>
> Regards,
> Bogdan
>
> On 06/24/2011 06:13 PM, Max Mühlbronner wrote:
>> Hi,
>>
>> yes, i tried several times. I should have mentioned the dr_rules has
>> quite a few (about 100k entries). I did notice it on a production
>> server running 1.6.2 but also was able to replicate the same behavior
>> when testing with the same Database/dr_rules on 1.6.4-tls.
>>
>> I will try to take a deeper look into the issue, will try to send
>> some traces / logs if there is something which looks like it could be
>> related.
>>
>>
>> BR
>>
>> Max M.
>>
>>
>> Am 24.06.2011 16:40, schrieb Bogdan-Andrei Iancu:
>>> Hi Max,
>>>
>>> the reload is done in a separate process and only when done, the
>>> routing data (in mem) is swapped, so while loading new data,
>>> opensips will still use the old one - it should not block and wait
>>> for reload.
>>>
>>> Are you sure is there any blocking  ? What version are you using ?
>>>
>>> Regards,
>>> Bogdan
>>>
>>>
>>>
>>> On 06/16/2011 03:48 PM, Max Mühlbronner wrote:
>>>> Hello,
>>>>
>>>>
>>>>
>>>> Opensips seems to not route my requests while reloading the
>>>> drouting rules from Database. Probably the DB operations are
>>>> blocking the remaining operations?
>>>>
>>>> Any idea if this is normal behavior, or misconfiguration on my
>>>> side? Does anyone know a solution for reloading while still
>>>> receiving traffic, so if the rules become quite big it will not
>>>> take the service down while reloading via mi (fifo dr_reload /
>>>> webinterface..)?
>>>>
>>>>
>>>> Best Regards
>>>>
>>>> Max M.
>>>>
>


_______________________________________________
Users mailing list
[hidden email]
http://lists.opensips.org/cgi-bin/mailman/listinfo/users
Reply | Threaded
Open this post in threaded view
|

Re: drouting - dr_reload (performance)

Bogdan-Andrei Iancu-2
Hi Max,

It may be related to syslog performance  -  check if you have the async
written enabled in syslog for the the opensips file.

Regards,
Bogdan

On 06/27/2011 03:40 PM, Max Mühlbronner wrote:

> Hi,
>
>
> I just found out what caused these issues, simple solution, thanks for
> your hints!
>
> If debug level is set to 0 it works just fine  + reload is really fast
> (wether set via fifo debug or in config). At level3 you can see lots
> of debugging messages when reloading, also the load goes up and the
> blocking occurs.
>
> At debug level 0 reload goes really fast and no blocking at all. So it
> really seems to be related to the xlogs/debugging.
>
> Everything is working fine for me now (at debug0). Thanks again.
>
> BR
>
> Max M.
>
> Am 25.06.2011 13:54, schrieb Bogdan-Andrei Iancu:
>> Hi Max,
>>
>> 100K rule is nothing for the dr module (disregarding the how fast the
>> DB is or not).
>>
>> Maybe the so called locking is about overall load or DB load - using
>> some xlogs and benchmarking module, you can try to see if there is an
>> actual blocking in the do_routing() function.
>>
>> Regards,
>> Bogdan
>>
>> On 06/24/2011 06:13 PM, Max Mühlbronner wrote:
>>> Hi,
>>>
>>> yes, i tried several times. I should have mentioned the dr_rules has
>>> quite a few (about 100k entries). I did notice it on a production
>>> server running 1.6.2 but also was able to replicate the same
>>> behavior when testing with the same Database/dr_rules on 1.6.4-tls.
>>>
>>> I will try to take a deeper look into the issue, will try to send
>>> some traces / logs if there is something which looks like it could
>>> be related.
>>>
>>>
>>> BR
>>>
>>> Max M.
>>>
>>>
>>> Am 24.06.2011 16:40, schrieb Bogdan-Andrei Iancu:
>>>> Hi Max,
>>>>
>>>> the reload is done in a separate process and only when done, the
>>>> routing data (in mem) is swapped, so while loading new data,
>>>> opensips will still use the old one - it should not block and wait
>>>> for reload.
>>>>
>>>> Are you sure is there any blocking  ? What version are you using ?
>>>>
>>>> Regards,
>>>> Bogdan
>>>>
>>>>
>>>>
>>>> On 06/16/2011 03:48 PM, Max Mühlbronner wrote:
>>>>> Hello,
>>>>>
>>>>>
>>>>>
>>>>> Opensips seems to not route my requests while reloading the
>>>>> drouting rules from Database. Probably the DB operations are
>>>>> blocking the remaining operations?
>>>>>
>>>>> Any idea if this is normal behavior, or misconfiguration on my
>>>>> side? Does anyone know a solution for reloading while still
>>>>> receiving traffic, so if the rules become quite big it will not
>>>>> take the service down while reloading via mi (fifo dr_reload /
>>>>> webinterface..)?
>>>>>
>>>>>
>>>>> Best Regards
>>>>>
>>>>> Max M.
>>>>>
>>
>
>
> _______________________________________________
> Users mailing list
> [hidden email]
> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>


--
Bogdan-Andrei Iancu
OpenSIPS solutions and "know-how"


_______________________________________________
Users mailing list
[hidden email]
http://lists.opensips.org/cgi-bin/mailman/listinfo/users