[Bucardo-general] Restarting replication after Postgresql restart.
Ben Allen
bsa at lanl.gov
Mon Apr 13 16:38:14 UTC 2009
No you didn't miss anything. Greg hasn't tagged a release since 3.0.7
for whatever reason, so what I'm saying is literally use the latest
development version. I know that doesn't really sound "right" but I'd
say the current code is stable, but obviously I would test it in your
environment fully before putting into production. The other option
would be to get Greg to tag a release that you could stay at.
Ben
On Apr 9, 2009, at 4:47 PM, Lev Tannen wrote:
> Hi Ben,
> Thank you for explanation.
> I feel I ask too much. But the more I learn the less I understand.
>
> I run git clone. I run "git tag -l" and got only two tags 3.0.6 and
> 3.0.7. I run "git branch" and received only "* master", I run "git
> remote" and got "original". In one of the previous emails you
> suggested to use the version 3.2.5. But I do not see even the tag
> 3.0.9. Where is it? I do not think it is correct to use a
> development version in production. I expect to see something named
> like "last stable version", "beta version" and so on. But there is
> nothing close.
> Did I missed something?
>
> Thank you,
> Lev
>
>
> On Thu, Apr 9, 2009 at 2:46 PM, Ben Allen <bsa at lanl.gov> wrote:
> Running "git clone http://bucardo.org/bucardo.git" will clone the
> repository to bucardo in the current directory. This bucardo folder
> will contain the latest code. You can simply use it as you have used
> the downloaded version. If you see new updates being pushed to the
> repository, run "git pull" inside the bucardo folder. This will
> download the latest version. If you've made any changes to your
> local copy it will try to merge those changes. If it fails, you'll
> have to manually merge the files. There's also the concept of
> committing changes you make locally. Running "git commit -a" will
> save your changes just to your local clone of the repo. For more
> info checkout: http://git-scm.com/documentation
>
> Ben
>
> On Apr 7, 2009, at 4:26 PM, Lev Tannen wrote:
>
>> Hi Ben,
>>
>> I finally figured out what was wrong. It was a verb "forcestart"
>> that Bucardo.pm used to restart itself that caused athe problem.
>> Bucardo_ctl did not recognize that verb. When I changed it to
>> "start" everything went in place.
>>
>> I think I have to switch to a version newer then 3.0.9. The problem
>> is it is the only version I could find. The article in http://bucardo.org
>> references this version. It also references a GIT repository, but
>> that GIT repository contains only the master version. So I am
>> completely lost.
>> Could you please provide me with information on how to access the
>> Bucardo repository, so I could see what versions exists, what is
>> the latest stable version and so on.
>> Thank you,
>> Lev
>>
>> On Tue, Apr 7, 2009 at 10:21 AM, Lev Tannen <levtannen at gmail.com>
>> wrote:
>> Hi Ben,
>> I run Bucardo as ./bucardo_ctl from the /home/bucardo/bucardo3-0-9
>> directory.
>>
>>
>> About the version 3-0-9, I use it because it is the only version
>> referred on bucardo.org.
>> I am not familiar with GIT. I have installed GIT on my computer and
>> issued a command git clone http://bucardo.org/bucardo.git, but what
>> has been downloaded looks just as one version of Bucardo and I do
>> not know which one. "git branch" lists only one branch: * master.
>> Could you please refer me to some manual so I will be able to use
>> git more productively.
>>
>> Thank you,
>> Lev
>> http://bucardo.org/bucardo.git
>>
>> On Mon, Apr 6, 2009 at 5:31 PM, Ben Allen <bsa at lanl.gov> wrote:
>> Hi Lev,
>>
>> Do you run bucardo_ctl when you are in the /home/bucardo/
>> bucardo3-0-9 directory? So "./bucardo_ctl", or do you run it like
>> "/home/bucardo/bucardo3-0-9/bucardo_ctl"
>>
>> As long as the bucardo user can access the database with however
>> you have setup permissions to the database there should be no
>> problem. Essentially all that bucardo does to restart itself after
>> a database failure is call bucardo_ctl start from itself.
>>
>> I don't see anything wrong with the bucardo.pm you attached. Have
>> you considered using the latest checkout version? Bucardo 3.0.9 is
>> quite old. The current version in Git is 3.2.5.
>>
>> Ben
>>
>> On Apr 6, 2009, at 3:53 PM, Lev Tannen wrote:
>>
>>> Ben,
>>>
>>> I successfully applied a patch, but it looks it does not work.
>>> When I restarted postgres, bucardo_ctl issued an error:
>>>
>>> [bucardo at devdb2 logs]$ FATAL: terminating connection due to
>>> administrator command
>>> Sync "Attempting automatic respawn after MCP death" does not
>>> appear to exist
>>>
>>> The bucardo log file gave me practically the same message:
>>> Sync "Attempting automatic respawn after MCP death"
>>>
>>> It waited for several minutes and restored the normal operation
>>> except for lucking direct respond to update. When I restarted
>>> bucardo (bucardo_ctl stop/start) direct respond has been restored.
>>>
>>> I see two possibilities: either I applied the patch incorrectly,
>>> or the patch does not work in my case.
>>> I have one complicating factor: I run bucardo_ctl from the /home/
>>> bucardo/bucardo3-0-9 directory under the user bucardo, but
>>> Postgresql is run from /usr/local/pgsql/bin directory under the
>>> user postgres. Could this explain why bucardo cannot restore a
>>> normal operation?
>>>
>>> I am attaching the patched bucardo.pm file.
>>>
>>> Thank you for any advise.
>>> Lev Tannen
>>>
>>>
>>>
>>> On Mon, Apr 6, 2009 at 2:33 PM, Ben Allen <bsa at lanl.gov> wrote:
>>> Try the attached patch. I made it based on Bucardo.pm from 3.0.9.
>>> I have not tested it though.
>>>
>>> Ben
>>>
>>>
>>> On Apr 6, 2009, at 11:15 AM, Lev Tannen wrote:
>>>
>>>> Ben,
>>>>
>>>> Thank you for advice. I tried to apply the patch, but received
>>>> error:
>>>>
>>>> [bucardo at devdb2 Bucardo-3.0.9]$ patch <patchText
>>>> patching file Bucardo.pm
>>>> patch: **** malformed patch at line 7: ## We assume this is
>>>> bucardo_ctl, and that we are in same directory as when called
>>>>
>>>> What would you suggest? I am running version Bucardo-3.0.9
>>>>
>>>> Thank you.
>>>> Lev
>>>>
>>>>
>>>> On Sat, Apr 4, 2009 at 5:03 PM, Ben Allen <bsa at lanl.gov> wrote:
>>>> Lev I'm pretty sure you are experiencing the same problem I was,
>>>> which was fixed by this patch. Basically Bucardo uses the
>>>> bucardo_ctl script itself to restart itself when there is a db
>>>> problem. The problem arises that Bucardo wasn't smart enough to
>>>> figure out the path of bucardo_ctl and assumed ./bucardo_ctl
>>>> which in most cases was wrong.
>>>>
>>>> If you are using 3.0.9 or version before this patch was committed
>>>> I wager to guess this is the problem.
>>>>
>>>> Regards,
>>>>
>>>> Ben
>>>>
>>>> Begin forwarded message:
>>>>
>>>>> From: bucardo-commits at bucardo.org
>>>>> Date: March 10, 2009 1:38:57 PM EDT
>>>>> To: bucardo-commits at bucardo.org
>>>>> Subject: [Bucardo-commits] [Bucardo] Check for executableness on
>>>>> MCP restart. Thanks to Ben Allen for the patch.
>>>>>
>>>>> Committed by Greg Sabino Mullane <greg at endpoint.com>
>>>>>
>>>>> Check for executableness on MCP restart. Thanks to Ben Allen for
>>>>> the patch.
>>>>>
>>>>> ---
>>>>> Bucardo.pm | 5 ++++-
>>>>> 1 files changed, 4 insertions(+), 1 deletions(-)
>>>>>
>>>>> diff --git a/Bucardo.pm b/Bucardo.pm
>>>>> index 5c39e68..ad855bf 100644
>>>>> --- a/Bucardo.pm
>>>>> +++ b/Bucardo.pm
>>>>> @@ -1158,7 +1158,10 @@ sub start_mcp {
>>>>>
>>>>> ## We assume this is bucardo_ctl, and that we are in same
>>>>> directory as when called
>>>>> my $RUNME = $old0;
>>>>> - $RUNME = "./$RUNME" if index($RUNME,'.')!=0;
>>>>> + ## Check to see if $RUNME is executable as is, before we
>>>>> assume we're in the same directory
>>>>> + if (! -x $RUNME) {
>>>>> + $RUNME = "./$RUNME" if index ($RUNME,'.') != 0;
>>>>> + }
>>>>> $RUNME .= qq{ start "Attempting automatic respawn after MCP
>>>>> death"};
>>>>> $self->glog("Respawn attempt: $RUNME");
>>>>> exec $RUNME;
>>>>> --
>>>>> 1.6.0.5
>>>>>
>>>>> _______________________________________________
>>>>> Bucardo-commits mailing list
>>>>> Bucardo-commits at bucardo.org
>>>>> https://mail.endcrypt.com/mailman/listinfo/bucardo-commits
>>>>
>>>>
>>>
>>>
>>> _______________________________________________
>>> Bucardo-general mailing list
>>> Bucardo-general at bucardo.org
>>> https://mail.endcrypt.com/mailman/listinfo/bucardo-general
>>>
>>>
>>> <Bucardo.pm>
>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://mail.endcrypt.com/pipermail/bucardo-general/attachments/20090413/a651a6cf/attachment.html
More information about the Bucardo-general
mailing list