Dimitri Fontaine wrote:
> Mark Kirkwood <mark.kirkwood@catalyst.net.nz> writes:
>
>> I've been using the wiki page
>> (http://wiki.postgresql.org/wiki/Streaming_Replication) as a guide, and I
>> notice that it recommends the master (and replicas) have a non-trivial
>> archive_command even after the backup step is completed. ISTM that after the
>> backup the master's archive_command can be set to '' or '/bin/true' as the
>> walsender does not make any use of the WAL archive (AFAICS anyway). Clearly
>> it might be desirable to have the archived segments around for other reasons
>> - but equally it might be desirable *not* to have to have to (e.g disk
>> space), or am I overlooking something?
>>
>
> I think it's still necessary in the case the connection between a slave
> and the master breaks. If the transient error is long enough for the
> slave requesting WALs that the master no longer has, restore_command
> will get used.
>
> IIUC from the mails here, the restore_command is run directly by the
> slave itself, so it needs to have access to master archives embedded in
> the restore_command.
>
> Take all this with a huge grain of salt, that's my understanding without
> having had the time to read the patch or play with it yet.
>
>
Thanks Dimitri, I'd missed that thread. Ok, slave will need a suitable
restore_comand in addition to primary_conninfo in recovery.conf, and
then extended communication failures (or shutting down the slave for a
while!) will not break the streaming setup (FWIW I tried this just now).
regards
Mark