[ZODB-Dev] plone and postgres "connection problems"

[lista administracion]

Thanks for the link and answer.

At the moment did tunning ¬ ¬! , Well better said a trial and error. I have
this configuration seems to work at least take a day without fail.


I leave my settings if anyone would work for more days.

Postgres with 4gigas
#####Set this to 140% of the average number of expected connections. 100
clients average means 140 max connections.
max_connections = 2000
tcp_keepalives_idle = 60                        # TCP_KEEPIDLE, in seconds;
                                        # 0 selects the system default
tcp_keepalives_interval = 30            # TCP_KEEPINTVL, in seconds;
                                        # 0 selects the system default
tcp_keepalives_count = 10               # TCP_KEEPCNT;
                                        # 0 selects the system default

shared_buffers = 262144                 # min 16 or max_connections*2, 8KB
each
#2 Gigas
#echo $((1024 * 1024 * 3072)) > /proc/sys/kernel/shmmax
#262144=1024*1024*3072/8192
work_mem = 16384
max_fsm_pages = 153600                  # min max_fsm_relations*16, 6 bytes
each
#GuyThornley's favourite settings for these parameters are:
commit_delay = 20000
commit_siblings = 3
effective_cache_size = 393216
random_page_cost = 4                    # units are one sequential page
fetch


To Plone 4 gigas
      zserver-threads 6
%import relstorage
    <relstorage>
        name BDplone
        blob-dir var/blobs
        cache-local-mb 512
        cache-delta-size-limit 10000
        commit-lock-timeout 30
        poll-interval 0
        pack-dry-run true
        pack-batch-timeout 5.0
        pack-duty-cycle 1.0
        pack-max-delay 20
        <postgresql>
           dsn dbname='BDplone' user='admin' host='10.9.10.33'
password='123'
        </postgresql>
    </relstorage>
    mount-point /
</zodb_db>




2010/4/19 Laurence Rowe <l at lrowe.co.uk>

> I've had this issue reported to me in the context of zope.sqlalchemy,
> but have been unable to reproduce it. Others have also seen it, but as
> far as I am aware have not been able to reproduce it:
> http://www.mail-archive.com/pgsql-hackers@postgresql.org/msg146522.html
>
> As there have now been three sightings, in the context of SQLAlchemy,
> Django, and plain dbapi2 usage (RelStorage). I suspect it is a real
> issue with Psycopg2, however until it can be reproduced I'm not
> hopeful it can be fixed.
>
> Laurence
>
> On 19 April 2010 19:34, lista administracion <reference.list at gmail.com>
> wrote:
> > Hi
> >
> > We have two servers with Plone that point to the same database [Postgres
> 8]
> > so that if one fails the other working.
> >
> > The problem is that at least once a day left to answer my Plone few
> minutes
> > (it is in white or takes a long to respond)
> > and cause apache send Proxy Error, this is solved but it is only annoying
> > for we users to wait 5 to 10 minutes not counting the lack of
> availability
> > of the page.
> >
> > We find that when it fails, PostgreSQL maintains a connection and keeps a
> > long time as "in transaction" and after about 10 minutes this connection
> > automatically terminates. this causes plone fail,at least 10 minutes
> >
> > The solution restart plone or kill the process as follows
> >
> > postgres 23267  0.0  0.1 2172016 7768 ?        S    10:17   0:00
> postgres:
> > ploneadmin plonetesting 10.9.33.116(45189) idle in transaction
> > kill -15 23267
> >
> >
> > We have
> > 2 Servers  with Apache/Plone
> > Ram           4 Gigas
> > RedHat        5.4
> > Apache        2.2.3
> > Plone         3.3.4
> > RelStorage-1.4.0b3
> >
> > 1 Server with Postgres
> > Ram           4 Gigas
> > RedHat        5.4
> > Postgres      8.1.18
> >
> > Configuration.
> >
> > -- Conexion DB
> >    <relstorage>
> >         blob-dir var/blobs
> >         <postgresql>
> >             dsn dbname='plone' user='admin' host='10.9.33.128'
> > password='password123'
> >         </postgresql>
> >     </relstorage>
> >
> > Recently modify these parameters in both servers and now it takes about 3
> > days to fail.
> >
> >         blob-dir var/blobs
> >         cache-local-mb 512
> >         cache-prefix prod
> >         cache-delta-size-limit 5000
> >         commit-lock-timeout 10
> >         poll-interval 60
> >         pack-dry-run true
> >         pack-batch-timeout 8
> >         pack-duty-cycle 0.3
> >         pack-max-delay 30
> >
> > We do not know if the problem is plone, postgres or network
> > I hope someone can advise something.
> >
> > thanks in advance.
> >
> > _______________________________________________
> > For more information about ZODB, see the ZODB Wiki:
> > http://www.zope.org/Wikis/ZODB/
> >
> > ZODB-Dev mailing list  -  ZODB-Dev at zope.org
> > https://mail.zope.org/mailman/listinfo/zodb-dev
> >
> >
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.zope.org/pipermail/zodb-dev/attachments/20100423/822dcc16/attachment.html