agent to fileserver retry
9 views
Skip to first unread message
asq
Nov 27, 2012, 11:55:55 AM11/27/12
to puppet...@googlegroups.com
hi,
we use active/standby load balancer in front of our puppetmaster cluster.
once in a while load banalcer vIP migrates (ie. when we reboot active node), which gives about 5-10 seconds of unavailability (until loadbalancer reconnects to backends). during that time it emits 503...
i wonder if it would be both useful and not hard to code to add a configuration option which alters agent so that its http client can retry (with configurable retry number, timeout and exponent). in our case it should just resume after those 10 seconds and apply the remainder of catalog, it might be even more important if you run puppetmaster over wan/vpn.
in current case it gives us some apply failures, which upsets our monitoring and graphing services...
what do you think?
we use active/standby load balancer in front of our puppetmaster cluster.
once in a while load banalcer vIP migrates (ie. when we reboot active node), which gives about 5-10 seconds of unavailability (until loadbalancer reconnects to backends). during that time it emits 503...
i wonder if it would be both useful and not hard to code to add a configuration option which alters agent so that its http client can retry (with configurable retry number, timeout and exponent). in our case it should just resume after those 10 seconds and apply the remainder of catalog, it might be even more important if you run puppetmaster over wan/vpn.
in current case it gives us some apply failures, which upsets our monitoring and graphing services...
what do you think?
Reply all
Reply to author
Forward
0 new messages