Hi Guys,
We have some pretty large lb_virtuals (12000 entries), lb_pools (180000 entries) tables. I found that the poller was struggling to keep up with this (understandable IMHO). What I did find was MySQL appeared to be the bottleneck so after some debugging I added the following indexes:
create index lb_virtuals_idx_1 on lb_virtuals (device_id,virt_name); create index lb_pools_idx_1 on lb_pools (device_id,pool_name); create index lb_snatpools_idx_1 on lb_snatpools (device_id,snatpool_name);
The middle one had the biggest impact, since that is the biggest table. Here is the polling result showing the difference:
Time Duration 2021-04-29 17:05:02 56s 2021-04-29 17:00:02 4m 19s
Hopefully this can help someone else if they have an issue like this (I hope not for your sanity :D ) or perhaps the devs can advise a better way of doing it :)
P.S. - The issue was that polling kept running over each other and that cause it to spiral out of control, data gaps where getting bigger and bigger and MySQL was pinned at 250-300% CPU (4 core box this)
Cheers,
Pieter