Hi all,
We’re using mongo 4.2.6 with a 4 members replica set, on a 24/7 running platform, without much usage spike during a regular day.
The platform has been in production without too many hiccups for about a year, but since 2 weeks we have seen a major CPU spike (load going from an usual 5 to 48) on the 48 cores primary server that happens everyday starting at midnight 12pm (UTC) and is either (no clear frequency, every 2 or 3 days) very short or lasts exactly one hour (1am loads drops almost instantly).
We have a backup running at another time and using readpreference to secondary so as not to charge the primary, so that’s ruled out, and no particular other processes that would start at 12pm.
We were wondering if there was some default maintenance processes that we might have overlooked that could cause the issue ?
Or any ideas on how to investigate the problem ?
Thanks a bundle,
Philippe