Обсуждение: possible memory leak in Server Status window

Поиск
Список
Период
Сортировка

possible memory leak in Server Status window

От
Julius Tuskenis
Дата:
Hello, all!

Last Friday I've installed pgAdmin 1.12.2 (build Dec 13, 2010) into one
of our clients servers (Windows server 2003 SE SP2). Postgresql 8.3.3
runs on it. As we were looking how to improve performance we were using
Server Status tool (great tool, especially with recent changes  - thank
you for it). I have noticed that pgAdmin consumes virtual memory when
the tool is running. The increment is small (~50Kb) and it happens every
6 seconds or so.. I have noticed, that even if I increase the update
interval to 1minute that does not effect the MV increase - it still
happens as before.
Today I have opened the Server Status tool, and during an hour or so the
VM size used by pgAdmin increased from 13Mb to 28Mb.
I tried to reproduce the behavior on other systems but with no success
yet. Could you give me a hint what I should try?

--
Julius Tuskenis
Programavimo skyriaus vadovas
UAB nSoft
mob. +37068233050


Re: possible memory leak in Server Status window

От
Guillaume Lelarge
Дата:
Le 07/02/2011 09:31, Julius Tuskenis a écrit :
> Hello, all!
>
> Last Friday I've installed pgAdmin 1.12.2 (build Dec 13, 2010) into one
> of our clients servers (Windows server 2003 SE SP2). Postgresql 8.3.3
> runs on it. As we were looking how to improve performance we were using
> Server Status tool (great tool, especially with recent changes  - thank
> you for it). I have noticed that pgAdmin consumes virtual memory when
> the tool is running. The increment is small (~50Kb) and it happens every
> 6 seconds or so.. I have noticed, that even if I increase the update
> interval to 1minute that does not effect the MV increase - it still
> happens as before.
> Today I have opened the Server Status tool, and during an hour or so the
> VM size used by pgAdmin increased from 13Mb to 28Mb.
> I tried to reproduce the behavior on other systems but with no success
> yet. Could you give me a hint what I should try?
>

Did you lower each report? because each individual report has its own timer.

I have never met this kind of issue with the server status window, and
noone complained yet about this.


--
Guillaume
 http://www.postgresql.fr
 http://dalibo.com

Re: possible memory leak in Server Status window

От
Julius Tuskenis
Дата:
2011.02.07 11:52, Guillaume Lelarge rašė:
> Did you lower each report? because each individual report has its own
> timer.
>
I'm sorry, I have only one report (Server status window). When you say
"lower" do you mean "minimize" ? No, I have left it on screen. Or do you
mean refresh interval? As I have single server status window, there is
no window left with shorter interval.

Could the answer be in Locks window? I noticed, that even when I set
refresh interval to 10min, lock list is updated after the interval of
approximately 1 minute. Is this expected behavior? And when having
refresh interval = 10min the events of increase to VM become rare. If I
hit refresh button - i can observe the increase.

I still was not able to reproduce on other servers. Maybe it is related
to the error we found Friday night in System log "The server was unable
to allocate from the system nonpaged pool because the pool was empty." I
tried to google - found an explanation that this could mean there is a
memory leak in kernel or in drivers. Maybe something is wrong in
system... This is the first time I encounter such a problem.

--
Julius Tuskenis
Programavimo skyriaus vadovas
UAB nSoft
mob. +37068233050


Re: possible memory leak in Server Status window

От
Julius Tuskenis
Дата:
2011.02.07 13:14, Julius Tuskenis rašė:
> 2011.02.07 11:52, Guillaume Lelarge rašė:
>> Did you lower each report? because each individual report has its own
>> timer.
>>
> I'm sorry, I have only one report (Server status window). When you say
> "lower" do you mean "minimize" ? No, I have left it on screen. Or do
> you mean refresh interval? As I have single server status window,
> there is no window left with shorter interval.
>
> Could the answer be in Locks window? I noticed, that even when I set
> refresh interval to 10min, lock list is updated after the interval of
> approximately 1 minute. Is this expected behavior? And when having
> refresh interval = 10min the events of increase to VM become rare. If
> I hit refresh button - i can observe the increase.
>
> I still was not able to reproduce on other servers. Maybe it is
> related to the error we found Friday night in System log "The server
> was unable to allocate from the system nonpaged pool because the pool
> was empty." I tried to google - found an explanation that this could
> mean there is a memory leak in kernel or in drivers. Maybe something
> is wrong in system... This is the first time I encounter such a problem.
>
I'm sorry - I guess I should ask for help in pgAdmin mailing list.

--
Julius Tuskenis
Programavimo skyriaus vadovas
UAB nSoft
mob. +37068233050


Re: possible memory leak in Server Status window

От
Guillaume Lelarge
Дата:
Le 07/02/2011 12:14, Julius Tuskenis a écrit :
> 2011.02.07 11:52, Guillaume Lelarge rašė:
>> Did you lower each report? because each individual report has its own
>> timer.
>>
> I'm sorry, I have only one report (Server status window). When you say
> "lower" do you mean "minimize" ? No, I have left it on screen. Or do you
> mean refresh interval? As I have single server status window, there is
> no window left with shorter interval.
>

There are a number of reports in the server status window. Each has
their own refresh interval.

> Could the answer be in Locks window? I noticed, that even when I set
> refresh interval to 10min, lock list is updated after the interval of
> approximately 1 minute. Is this expected behavior? And when having
> refresh interval = 10min the events of increase to VM become rare. If I
> hit refresh button - i can observe the increase.
>

You need to set specifically its refresh interval. But you may have
found the issue. I still need to look at the code.

> I still was not able to reproduce on other servers. Maybe it is related
> to the error we found Friday night in System log "The server was unable
> to allocate from the system nonpaged pool because the pool was empty." I
> tried to google - found an explanation that this could mean there is a
> memory leak in kernel or in drivers. Maybe something is wrong in
> system... This is the first time I encounter such a problem.
>

So do I.


--
Guillaume
 http://www.postgresql.fr
 http://dalibo.com