On 10/18/2011 09:44 AM, Simon Riggs wrote:
> On Tue, Oct 18, 2011 at 4:58 PM, David Kerr<dmk@mr-paradox.net> wrote:
>
>> I have postgres setup for streaming replication and my slave box went down.
>>
>> My question is, how long can that box stay down before it causes a material impact on the master?
>>
>> The archive_command that I use will not archive logs while the slave is down.
>>
>> I know the obvious problems:
>> * you're not archiving, so you lose all of the recovery related benefits if you should crash
>> * could run out of disk space in pg_xlog
>> * your slave could get out of sync
>>
>> What i'm concerned with is something like, if the master is unable to archive eventually
>> it will stop writing WALs Or something impacting the performance of the database.
>> or anything along those lines.
>
>
> When it comes back up it will have to catchup. At some point it will
> be quicker to regenerate the standby than to catchup.
>
> Also, at some point you will run out of space in pg_xlog, which would
> make the master crash. So probably best to have an archive_command
> that starts deleting or compressing files before disk fills, but that
> means your slave can then never catch up at that point.
>
Ok, that's better that I thought. Thanks!
