Hi all, Is that possible to start pollers with the delay? like when you have 200+ clusters a delay between each poller (or a group of pollers) to spread the load. Can't find such a feature documented
#Starting pollers with the delay
lament field
rancid sphinx
hi @lament field there is not an option. Can you say more about which load you're trying to spread? Since there is one poller per cluster, the communication between pollers is already independent. The pollers don't communicate with each other and each poller is only talking to one cluster
lament field
the implementation has just started for 400+ ONTAP cluster nodes. The situation observed is when there is a number of pollers configured, some of them do not start and harvest restart fixes that. The suggestion was that maybe high number of parallel pollers somehow disturb network/compute resources and the delay could help to spread the load in time
rancid sphinx
thanks! If you get a chance, make sure that the status is not misleading. I suspect all the pollers are starting but the bin/harvest command that prints the status is missing some of them. Basically it's a race condition. Perhaps we can improve that printing. The way to confirm is by running start, the output shows the status of some as "not running". Run bin/harvest status and notice the pollers that were listed as not running are now running
lament field
can i monitor any specific message in the logs?
rancid sphinx
if you are interested in startup, probably easier to check bin/harvest status or systemctl status harvest or ps aux | grep poller Once the poller is created, the logs for that poller will be in /var/log/harvest by default