Hi all,
Our staging area has 3 PostgreSQL servers running streaming replication with WAL archiving in a primary with 2 standbys
arrangement.All of these are running on VMs with CentOS
6.5. Things were running fine until this morning when our QA folks ran a code update, and during the update, one of the
standbyservers died (I don't know why yet). The other 2
systems were fine so the primary could continue to stream to the remaining standby. but now some of the code is
intermittentlyfailing due to timeouts. Yes, I've asked the
developer to modify his code to allow some additional time for completion.
I've asked QA if they know whether a replication connection from the dead standby was active when it died, but they
haven'tanswered me yet. I also reminded them that had this been
a single standby, the system would now be completely unresponsive and not just slow. But my question really is, why is
theperformance degraded in this event? Seems to me that it
should be behaving as a primary with one standby and running normally.
Ideas?
--
Jay