Christopher Browne wrote:
> JP Fletcher <jpfletch at ca.afilias.info> writes:
>   
>> I'm experimenting with logshipping (v.1.2.11) and am using, for the
>> first time, the option in the slon conf file:
>>
>> command_on_logarchive
>>
>> I created a shell script that this command calls which applies syncs
>> archives after they are written:
>>
>>   command_on_logarchive="/opt/clusters/cluster1_slony/archive/apply_archive.sh"
>>
>> My cluster has four nodes, not counting the log shipped replica, 8141,
>> 8142, 8143, and 8194.  Node 8194 is the provider to the logshipped
>> node, so I only configured 'command_on_logarchive' for the slon
>> servicing this node, my assumption is that this is correct.  Nodes
>> 8143, and 8141 are origins for sets.
>>
>> When I started things up, everything appeared to be fine.  The slon
>> successfully applied 37 archives, each originating on either node 8141
>> or 8143.  The 38th archive, however, and several subsequent archives
>> originated on node 8142.  The slon made no attempt to apply these
>> archives.  This left the archive counter  on 37.  The next archive the
>> slon attempted to apply was from node 8143, and  expected the  archive
>> counter to be on 48, which of course it was not, and everything went
>> pear-shaped.   I have three questions:
>>
>> 1.  Should I have expected the command_on_logarchive option to work,
>> and if so, did i do something wrong?
>>     
>
> Hopefully apply_archive.sh looked something like:
>
> #!/bin/sh
> PGBINDIR=????
> ${PGBINDIR}/slony_logshipper $1
>
> That might be hoped to work, although not quite right yet in 1.2.11
>   

#!/bin/bash

SQL_ARCHIVE=$1

/opt/dbs/pgsql81_slonytest/bin/psql -h yrg-testbed2 -p 9533 -U slony -f 
$SQL_ARCHIVE cluster1_logship >> 
/opt/clusters/cluster1_slony/archive/apply_archive.log 2>&1

if [ $? != 0 ]; then

   echo "Failed to apply archive $SQL_ARCHIVE >> 
/opt/clusters/cluster1_slony/archive/apply_archive.log 2>&1"
   exit 1
  fi

mv $SQL_ARCHIVE $SQL_ARCHIVE.applied

> :-(.
>
>   
>> 2.  Why does the slon not try to apply archives from node8142?
>>     
>
> slon shouldn't; I think you're meaning the slony_logshipper...
>   

Yes, of course you're correct, but it doesn't look like the shell script 
was even run for these files.
>   
>> 3.  Why are there archives from a node that is not active (no events
>> should have originated on 8142, and no archive files exist for the
>> other non-origin node)?
>>     
>
> The node that isn't active should have very few, very infrequent,
> archives that are devoid of actual content (beyond the headers).
>
> It is expected that there *will* be some events coming from other
> nodes, as every node does generate events, even if infrequently.
>
> The behaviour will change a bit in 1.2.12; Jan just committed some
> changes at the end of last week to fix up the handling of sequencing
> of the archives, which turned out to be (sigh) troublesome.
>   


-- 
JP Fletcher
Database Administrator
Afilias Canada
voice: 416.646.3304 ext. 4123
fax: 416.646.3305
mobile: 416.561.4763
jpfletch at ca.afilias.info