On 8/29/2004 11:45 PM, Christopher Browne wrote:

> Ben said:
>> Hey guys,
>>
>> I'm trying to get slony going on my 7.4.2 postgres install on linux,
>> and I'm running into problems.
>>
>> Things seem to be going great until I start up slon on the slave. I see
>> this:
>>
>>
>> WARN   remoteWorker_wakeup: node 1 - no worker thread
>>
>>
>> Then, once I start replication, I see a lot of stuff like this on the
>> slave:
>>
>> WARN   remoteWorkerThread_1: transactions earlier than XID 109105155
>> are still in progress
>> WARN   remoteWorkerThread_1: data copy for set 1 failed - sleep 15
>> seconds
>>
>> I've read reports that libpq sometimes needs to be compiled with thread
>> safety turned on, but the same things say that's not the case on linux.
>> So.... thoughts?
> 
> That sounds somewhat consistent with the "threads issue," except that yes,
> that's not a Linux issue.  (It has bitten me on both Solaris and AIX :-(.)
> 
> You should take a peek at what transactions are open on the "master" DB. 
> Perhaps there's a pretty old connection open.

It's not the --enable-thread-safety, but exactly this one. The new 
subscriber cannot begin copying the data if transactions are still open 
that existed when the set origin enabled the subscription.

> 
> I have seen this, restarted the postmaster (forcibly terminating any O/S
> connections), and watched the copying proceed.

Restarting the application that holds open transactions indefinitely 
(often some bogus autocommit-off implementation attempt) should be enough.


Jan

> 
> I don't really like that "solution" any more than I like the notion of
> rebooting systems just because that seems to 'do the trick.'  But at any
> rate, take a look at what locks are open by doing "SELECT * FROM
> PG_LOCKS;"  That may tell you what old process is still hanging around.
> 
> There's a documentation file where it would be nice to drop docs on this
> so that we could, in future, point to "See item #17 on the 'typical
> problems' list."


-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #