| Puppet Version: 7.18.0
Puppet Server Version: 2021.7.0.16
OS Name/Version: Oracle Linux 7 When lowering runinterval on agents, from 60m to 30m for example, splay on, and splaylimit set to runinterval, agents quicker trigger a run and ignore the splay setting. We use the puppet_conf bolt task to make agent config changes, so a large group of agents suddenly synchronize and stay that way when we update their runintervals on disk. Restarting agents resolves the issue because the splay setting will be used. We run into this because sometimes we set runinterval to a higher value when we think we have performance problems on the compilers or primary server. Reverting the setting then causes this issue, which causes performance problems To reproduce: bolt task run puppet_conf action=set setting=splay section=agent value=true
bolt task run puppet_conf action=set setting=splaylimit section=agent value=10m
bolt task run puppet_conf action=set setting=runinterval section=agent value=10m They're synchronized now, the fix is restarting all the agents. That's not bad, but I don't see the benefit to having the agent reread the puppet.conf every few seconds if it can't handle an updated runinterval properly. Desired Behavior: I would expect it the agent schedule to work similarly to as if I had just restarted the agent. Actual Behavior: Agents stack up and run simultaneously. |
