Fwd: [MAPSERVER-USERS] CLOSE_CONNECTION=DEFER - Segmentation fault
Ivan Mincik
ivan.mincik at gmail.com
Mon Mar 17 13:42:21 EDT 2008
Hi,
how is the situation about the patch? Can I try some?
Thanks for Your time
Ivan
On Wednesday 12 March 2008 23:45, you wrote:
> Oops, forgot to CC mapserver-users.
>
> ---------- Forwarded message ----------
> From: Dave Fuhry <dfuhry at gmail.com>
> Date: Wed, Mar 12, 2008 at 6:33 PM
> Subject: Re: [MAPSERVER-USERS] CLOSE_CONNECTION=DEFER - Segmentation fault
> To: Ivan Mincik <ivan.mincik at gmail.com>
>
>
> Ivan,
>
> Ah, that does help. It looks like the error:
>
> ERROR: DECLARE CURSOR may only be used in transaction blocks
> is occurring after ROLLBACK statements are issued. I had not really
> considered ROLLBACKs, not having seen them in my tests.
>
> ROLLBACKs are only issued in two places in mappostgis.c. In
> msPOSTGISLayerGetShape(), and in prepare_database() (but in the latter
> case, only if using an ancient version of postgis).
>
> The code in msPOSTGISLayerGetShape() seems unnecessary to me; no other
> DECLARE CURSOR ...; FETCH ...; CLOSE ...; queries are followed by a
> ROLLBACK, so it seems unnecessary here as well. The important thing
> will be to ensure that if an error happens in one of those queries,
> that the connection is not msConnPoolRelease()d back into the
> connection pool. I haven't traced the code in detail yet, but that
> should be staightforward enforce.
>
> So essentially, the patch should be modified to remove the ROLLBACKs
> from msPOSTGISLayerGetShape() in a safe way.
>
> Thanks,
>
> Dave
>
> On Wed, Mar 12, 2008 at 5:49 AM, Ivan Mincik <ivan.mincik at gmail.com> wrote:
> > Hi Dave,
> > these are my postgis version details:
> > select postgis_full_version();
> > postgis_full_version
> >
> > ------------------------------------------------------------------------
> >---------- POSTGIS="1.3.1" GEOS="2.2.3-CAPI-1.1.1" PROJ="Rel. 4.5.0, 22
> > Oct 2006" USE_STATS
> > (1 row)
> >
> > 1. I have changed value log_min_duration_statement
> > 2. Restart of postgresql
> > 3. Restart of apache
> > 4. Restart of all firefox windows and deleted cache
> > 5. pmapper loaded
> >
> > Here are two logs when reproducing the problem:
> > http://projects.gpsmapy.sk/logs/
> >
> > I have tried to use mapserver also as fastcgi, but same problem. Every
> > time the problem will start when I try to identify the object
> > (Obj_queryBy{Point,Rect,etc.}).
> >
> > Thanks a lot
> > Ivan
> >
> > On Wednesday 12 March 2008 04:41, you wrote:
> > > Ivan,
> > >
> > > p.mapper seems to use the typical php
> > > {map,layer}Obj_queryBy{Point,Rect,etc.} functions. These calls end
> > > up calling the same msPOSTGIS* functions as everything else.
> > >
> > > I can't seem to reproduce your problem with a simple testcase
> > > (basically, the attached files). It might help if you can set
> > > log_min_duration_statement = 0 in your postgresql.conf, reproduce the
> > > problem, and send the output from your postgresql log file. It will
> > > show all commands issued to the database, including the BEGIN, and
> > > the DECLARE CURSOR ..., and FETCH for every layer.
> > >
> > > I can setup p.mapper if that's necessary to reproduce the problem.
> > >
> > > (By the way, I assume you're not using an ancient version of
> > > postgis, in the 0.5 - 0.6 era. There is an alternate codepath for
> > > that, which I have not tested.)
> > >
> > > Thanks,
> > >
> > > Dave
> > >
> > > On Sun, Mar 9, 2008 at 5:31 AM, Ivan Mincik <ivan.mincik at gmail.com>
wrote:
> > > > I am sorry for late reply,
> > > > I am testing mapserver with p.mapper application so I *don't* use
> > > > it as cgi/fcgi.
> > > >
> > > > I have downloaded source package of mapserver 5.0.2 from debian
> > > > unstable, applied a patch and compiled on Debian Etch.
> > > > Installed and restarted apache and also rebooted whole system few
> > > > times (I am testing this on my desktop computer).
> > > >
> > > > The problems are starting mostly when I am trying to use
> > > > p.mapper's identify function, to get attributes of map object. Then
> > > > p.mapper is freezing. In mapfile i have 7 postgis layers.
> > > >
> > > > I can try this also with some openlayers which is using cgi.
> > > >
> > > > LOGS:
> > > >
> > > > 2008-03-09 10:24:44 CET WARNING: there is already a transaction
> > > > in progress 2008-03-09 10:24:45 CET WARNING: there is already a
>
> transaction
>
> > > > in progress 2008-03-09 10:24:46 CET ERROR: DECLARE CURSOR may only
> > > > be used in transaction blocks
> > > > 2008-03-09 10:24:46 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:46 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:46 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:46 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:47 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:47 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:47 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:47 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:50 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:50 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:50 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:52 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:52 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:52 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:53 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:53 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:53 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:54 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:54 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:54 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:54 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:54 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:54 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:24:55 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:24:55 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:24:55 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:25:00 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:25:00 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:25:00 CET ERROR: cursor "mycursor" does not
> > > > exist 2008-03-09 10:25:00 CET ERROR: DECLARE CURSOR may only be
> > > > used in transaction blocks
> > > > 2008-03-09 10:25:00 CET WARNING: there is no transaction in
> > > > progress 2008-03-09 10:25:01 CET ERROR: cursor "mycursor" does not
> > > > exist
> > > >
> > > >
> > > > Thanks a lot
> > > > Ivan
> > > >
> > > > On Wednesday 05 March 2008 22:41, you wrote:
> > > > > Ivan,
> > > > >
> > > > > Thanks for the problem report. If you are using fastcgi, you
> > > > > may need to restart it.
> > > > >
> > > > > The only two places PQconnectdb is called is in
> > > > > msPOSTGISLayerOpen and in msPOSTGRESQLJoinConnect, which
> > > > > operates independently, not using cursors or the connection
> > > > > pool. With a patched
> > > > > msPOSTGISLayerOpen, PQexec(layerinfo->conn, "BEGIN") is always
> > > > > called right after connecting, before any queries and before the
> > > > > connection ends up in the connection pool.
> > > > >
> > > > > So I'm wondering if you've put the new mapserv executable in
> > > > > your cgi-bin, but not restarted the webserver, and it's pulling
> > > > > old connections out of the fastcgi connection pool, which have
> > > > > not had PQexec(layerinfo->conn, "BEGIN") issued to them (and
> > > > > thus, no transaction started).
> > > > >
> > > > > Try restarting fastcgi. Depending on your configuration,
> > > > > restarting your webserver may automatically restart fastcgi, or
> > > > > you may have to specifically restart the fastcgi service/daemon
> > > > > on that machine.
> > > > >
> > > > > Please let us know if this fixes the problem.
> > > > >
> > > > > Thanks,
> > > > >
> > > > > Dave
> > > > >
> > > > > On Wed, Mar 5, 2008 at 7:26 AM, Ivan Mincik
> > > > > <ivan.mincik at gmail.com>
> >
> > wrote:
> > > > > > Nothing is perfect. There are no problems with applied patch
> > > > > > when browsing the map. But when I use identify object query
> > > > > > from php5-mapscript the problem is rising again.
> > > > > >
> > > > > > I have in postgresql log:
> > > > > >
> > > > > > 2008-03-05 13:19:30 CET WARNING: there is no transaction in
> > > > > > progress 2008-03-05 13:19:30 CET ERROR: cursor "mycursor"
> > > > > > does not exist 2008-03-05 13:19:31 CET ERROR: DECLARE CURSOR
> > > > > > may only be used in transaction blocks
> > > > > > 2008-03-05 13:19:31 CET WARNING: there is no transaction in
> > > > > > progress 2008-03-05 13:19:31 CET ERROR: cursor "mycursor"
> > > > > > does not exist 2008-03-05 13:19:32 CET ERROR: DECLARE CURSOR
> > > > > > may only be used in transaction blocks
> > > > > > 2008-03-05 13:19:32 CET WARNING: there is no transaction in
> > > > > > progress 2008-03-05 13:19:32 CET ERROR: cursor "mycursor"
> > > > > > does not exist 2008-03-05 13:19:34 CET ERROR: DECLARE CURSOR
> > > > > > may only be used in transaction blocks
> > > > > > 2008-03-05 13:19:34 CET WARNING: there is no transaction in
> > > > > > progress 2008-03-05 13:19:34 CET ERROR: cursor "mycursor"
> > > > > > does not exist 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: autovacuum:
> > > > > > processing database "gisplan" 2008-03-05 13:20:10 CET LOG:
> > > > > > could not receive data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection 2008-03-05 13:20:10 CET LOG: could not receive
> > > > > > data from client: Connection reset by peer
> > > > > > 2008-03-05 13:20:10 CET LOG: unexpected EOF on client
> > > > > > connection
> > > > > >
> > > > > >
> > > > > > 2008-03-05 13:20:40 CET WARNING: there is already a
>
> transaction in
>
> > > > > > progress 2008-03-05 13:20:40 CET WARNING: there is already a
> > > > > > transaction in progress 2008-03-05 13:20:41 CET WARNING:
> > > > > > there is already a transaction in progress 2008-03-05 13:20:42
> > > > > > CET WARNING: there is already a transaction in progress
> > > > > > 2008-03-05 13:20:43 CET WARNING: there is already a
> > > > > > transaction in progress 2008-03-05 13:20:43 CET WARNING: there
> > > > > > is already a transaction in progress 2008-03-05 13:20:44 CET
> > > > > > WARNING: there is already a transaction in progress
> > > > > > 2008-03-05 13:20:44 CET WARNING: there is already a
> > > > > > transaction in progress 2008-03-05 13:20:45 CET WARNING:
> > > > > > there is already a transaction in progress 2008-03-05 13:20:45
> > > > > > CET WARNING: there is already a transaction in progress
> > > > > > 2008-03-05 13:20:45 CET WARNING: there is already a
> > > > > > transaction in progress 2008-03-05 13:20:46 CET WARNING:
> > > > > > there is already a transaction in progress 2008-03-05 13:20:46
> > > > > > CET WARNING: there is already a transaction in progress
> > > > > > 2008-03-05 13:20:46 CET WARNING: there is already a
> > > > > > transaction in progress 2008-03-05 13:20:47 CET WARNING: there
> > > > > > is already a transaction in progress 2008-03-05 13:20:47 CET
> > > > > > WARNING: there is already a transaction in progress 2008-03-05
> > > > > > 13:20:48 CET WARNING: there is already a transaction in
> > > > > > progress 2008-03-05 13:20:49 CET WARNING: there is already a
> > > > > > transaction in progress 2008-03-05 13:20:50 CET WARNING:
> > > > > > there is already a transaction in progress 2008-03-05 13:20:50
> > > > > > CET WARNING: there is already a transaction in progress
> > > > > >
> > > > > >
> > > > > > ivan
> > > > > >
> > > > > > On Monday 03 March 2008 21:53, rich.fromm wrote:
> > > > > > > hrach wrote:
> > > > > > > > I use mapserver 4.10.3, postgresql 8.1, Debian Etch.
> > > > > > > >
> > > > > > > > Is anybody using CLOSE_CONNECTION=DEFER without any
> > > > > > > > problems ?
> > > > > > >
> > > > > > > I am now using it fine with mapserver 5.0.0 after applying
> > > > > > > the patch for bug 2497:
> > > > > > > http://trac.osgeo.org/mapserver/ticket/2497
> > > > > > >
> > > > > > > I don't know if 4.10.3 is different enough for that patch
> > > > > > > to be applicable or not. But if it's not too hard to
> > > > > > > upgrade, you might want to try a later version (e.g. 5.0.2)
> > > > > > > plus that patch. Note that the patch may make it into the
> > > > > > > 5.0.3 release (it was originally targetted for 5.2).
> > > > > > >
> > > > > > > You also might want to take a look at the following related
> > > > > > > mailing list threads:
>
> http://www.nabble.com/Segmentation-fault---mapserver---postgis-to
>
> > > > > > >12653 945.h tml#a12674499
> > > > > > >
> > > > > > > http://www.nabble.com/postgres-transaction-warnings-to157493
> > > > > > >33.ht
> > > > > > >
> > > > > > >ml
> > > > > > >
> > > > > > > - Rich
> > > > > >
> > > > > > _______________________________________________
> > > > > > mapserver-users mailing list
> > > > > > mapserver-users at lists.osgeo.org
> > > > > > http://lists.osgeo.org/mailman/listinfo/mapserver-users
More information about the mapserver-users
mailing list