<?xml version="1.0" encoding="UTF-8" standalone="yes" ?>
<!DOCTYPE bugzilla SYSTEM "http://www.slony.info/bugzilla/bugzilla.dtd">

<bugzilla version="3.4.4"
          urlbase="http://www.slony.info/bugzilla/"
          
          maintainer="devrim@gunduz.org"
>

    <bug>
          <bug_id>166</bug_id>
          
          <creation_ts>2010-11-17 05:38:00 -0800</creation_ts>
          <short_desc>Provide a way of counting the number of outstanding operations in a sync</short_desc>
          <delta_ts>2010-11-17 09:51:40 -0800</delta_ts>
          <reporter_accessible>1</reporter_accessible>
          <cclist_accessible>1</cclist_accessible>
          <classification_id>1</classification_id>
          <classification>Unclassified</classification>
          <product>Slony-I</product>
          <component>core scripts</component>
          <version>devel</version>
          <rep_platform>All</rep_platform>
          <op_sys>All</op_sys>
          <bug_status>NEW</bug_status>
          
          
          
          
          
          
          <priority>low</priority>
          <bug_severity>enhancement</bug_severity>
          <target_milestone>---</target_milestone>
          
          
          
          <everconfirmed>1</everconfirmed>
          <reporter name="Steve Singer">ssinger</reporter>
          <assigned_to name="Slony Bugs List">slony1-bugs</assigned_to>
          <cc>slony1-bugs</cc>

      

      

      

          <long_desc isprivate="0">
            <who name="Steve Singer">ssinger</who>
            <bug_when>2010-11-17 05:38:55 -0800</bug_when>
            <thetext>This is a feature request.

It would be very useful from a &apos;what the heck is my system doing&apos; point of view to know how many operations make up each outstanding SYNC request.

A system that is behind by 1000 SYNC&apos;s with no row edits in any of them is in a very different state than a system that is 1000 SYNC&apos;s behind each with 10,000 UPDATE statements.

Similarly if the next/current SYNC to be processed consists of a million rows that might explain why replication &apos;seems&apos; to have stopped from an sl_status point of view.

The test_slony_state.pl script seems like the most logical place to add this though other ideas are welcome.</thetext>
          </long_desc>
          <long_desc isprivate="0">
            <who name="Christopher Browne">cbbrowne</who>
            <bug_when>2010-11-17 09:51:40 -0800</bug_when>
            <thetext>The logic for this is liable to need a stored function (or similar) which duplicates the logic in remote_worker.c which associates XIDs with SYNCs.

If we implement this (which could be quite a good idea!), this might provide a way to shift the SYNC contents evaluation out of sync_event() (in src/slon/remote_worker.c) into (a) stored function(s).

It&apos;s definitely nontrivial, but, if done, would pull a pretty complicated piece out of the C code.</thetext>
          </long_desc>
      
      

    </bug>

</bugzilla>