Thread: Re: [PATCHES] pg_standby
Hi Simon,<br /> I would preserve the existing trigger function as little t "-t", and maybe implement a catchup trigger functionas big t "-T"? Set it up so that if the first attempt to find the WAL file postgres is currently requesting succeeds,skip over the trigger check. If the first attempt fails, then do your trigger check. That way, in the OCF script,the postmaster can be started, the trigger file set, and connection to the database looped on until it succeeds asan indication for when the database is up and available. I think that's cleaner than comparing a filename from a 'ps' command.Once I've completed the OCF script and done some testing, I'll forward it to you for you to review and see if youwant to include it.<br /><br /> Thanks,<br /> Doug<br /><br /> On Thu, 2007-03-08 at 15:37 +0000, Simon Riggs wrote: <blockquotetype="CITE"><pre> <font color="#000000">On Thu, 2007-03-08 at 10:33 -0500, Doug Knight wrote:</font> <font color="#000000">> Thanks, Simon. I kind of figured that's how pg_standby would work,</font> <font color="#000000">> since its invoked by postgres once per WAL file. What I was thinking I</font> <font color="#000000">> might do in the OCF script is to grab the pg_standby process line from</font> <font color="#000000">> a ps, pull out the "current" WAL file path and filename, then do an</font> <font color="#000000">> existence check for the file. If the file exists, then</font> <font color="#000000">> pg_standby/postgres is probably processing it. If not, then we're</font> <font color="#000000">> probably waiting on it, implying that recovery is complete. Thoughts</font> <font color="#000000">> on this process?</font> <font color="#000000">I suppose I might be able to have the option to catch up before it</font> <font color="#000000">stops, on the basis that if it can find the file it was looking for</font> <font color="#000000">without waiting then that can override the trigger.</font> <font color="#000000">Which way would you like it to work?</font> </pre></blockquote>
On Thu, 2007-03-08 at 13:29 -0500, Doug Knight wrote: > I would preserve the existing trigger function as little t "-t", and > maybe implement a catchup trigger function as big t "-T"? Set it up so > that if the first attempt to find the WAL file postgres is currently > requesting succeeds, skip over the trigger check. If the first attempt > fails, then do your trigger check. That way, in the OCF script, the > postmaster can be started, the trigger file set, and connection to the > database looped on until it succeeds as an indication for when the > database is up and available. I think that's cleaner than comparing a > filename from a 'ps' command. Once I've completed the OCF script and > done some testing, I'll forward it to you for you to review and see if > you want to include it. I'm happy to do this, unless other objections. I'll be doing another version before feature freeze. -- Simon Riggs EnterpriseDB http://www.enterprisedb.com
Excellent. Once you're ready, fire it over and I'll test it on our config.<br /><br /> Doug<br /> On Thu, 2007-03-08 at18:34 +0000, Simon Riggs wrote: <blockquote type="CITE"><pre> <font color="#000000">On Thu, 2007-03-08 at 13:29 -0500, Doug Knight wrote:</font> <font color="#000000">> I would preserve the existing trigger function as little t "-t", and</font> <font color="#000000">> maybe implement a catchup trigger function as big t "-T"? Set it up so</font> <font color="#000000">> that if the first attempt to find the WAL file postgres is currently</font> <font color="#000000">> requesting succeeds, skip over the trigger check. If the first attempt</font> <font color="#000000">> fails, then do your trigger check. That way, in the OCF script, the</font> <font color="#000000">> postmaster can be started, the trigger file set, and connection to the</font> <font color="#000000">> database looped on until it succeeds as an indication for when the</font> <font color="#000000">> database is up and available. I think that's cleaner than comparing a</font> <font color="#000000">> filename from a 'ps' command. Once I've completed the OCF script and</font> <font color="#000000">> done some testing, I'll forward it to you for you to review and see if</font> <font color="#000000">> you want to include it.</font> <font color="#000000">I'm happy to do this, unless other objections.</font> <font color="#000000">I'll be doing another version before feature freeze.</font> </pre></blockquote>
Simon Riggs wrote: > On Thu, 2007-03-08 at 13:29 -0500, Doug Knight wrote: > > > I would preserve the existing trigger function as little t "-t", and > > maybe implement a catchup trigger function as big t "-T"? Set it up so > > that if the first attempt to find the WAL file postgres is currently > > requesting succeeds, skip over the trigger check. If the first attempt > > fails, then do your trigger check. That way, in the OCF script, the > > postmaster can be started, the trigger file set, and connection to the > > database looped on until it succeeds as an indication for when the > > database is up and available. I think that's cleaner than comparing a > > filename from a 'ps' command. Once I've completed the OCF script and > > done some testing, I'll forward it to you for you to review and see if > > you want to include it. > > I'm happy to do this, unless other objections. > > I'll be doing another version before feature freeze. Should we be getting a patch for this for 8.3? -- Bruce Momjian <bruce@momjian.us> http://momjian.us EnterpriseDB http://www.enterprisedb.com + If your life is a hard drive, Christ can be your backup. +