Hi
Also, the grace period stuff from v2.3 and on should make it possible to
restart pound very quickly (milliseconds of downtime), and still
giving established tcp sessions a fair chance to complete.
Tracked sessions will be lost, but IP based persistence should work after
a config reload for all groups where the backend server list isn't
modified. At least I believe so..
You could even upgrade pound with practically no downtime.
Just a suggestion, and possibly not a good one:
When restarting pound the newly started instance could connect to the old
one and retreive the session table using the poundctl mechanism.
It could also delay binding until the session table is copied, and the
closing/binding could be synced between the old and new instance through
the same interface for even shorter downtime.
Rune
---
Rune Sætre <rune.saetre(at)netcom-gsm.no>
NetCom as
..
On Wed, 23 May 2007, Ondra Kudlik wrote:
> Hello,
>
> just try to search through archive, it is very frequent question.
> Current solution i.e.:
>
>
http://www.apsis.ch/pound/pound_list/archive/2005/2005-08/1125362245000/index_html?fullMode=1#1125499291000
>
> I just implemented this solution and with small tweaks it is fully
> functional.
>
> Problem is still when you need to upgrade "master" pound or if you
> need to listen to new https (new ip...)
>
> Kepi
>
> Út, kvÿÿ 22, 2007 ve 09:02:10 +0200, Francisco Ruiz napsal:
>> Hello,
>>
>> I would like to know if there is the intention of implementing reloading
>> the configuration via a signal HUP to the proccess. I think this will be
>> a very usefull feature for many people.
>>
>> Regards.
>
> --
> To unsubscribe send an email with subject 'unsubscribe' to pound(at)apsis.ch.
> Please contact roseg(at)apsis.ch for questions.
>
http://www.apsis.ch/pound/pound_list/archive/2007/2007-05/1179817330000/1179909469000
>
|
