HI,
I am currently facing an interesting issue. Instance M10, 6k documents in one database. Exporting bson using mongodump. Export is OK, BUT after it, Atlas starts to behave “strange”
CPU usage on P node stays on 30% (normally 3-10%), and network outgoing traffic remains at 2-3MB/s (normally 100-400 kb/s), despite the fact, that export is finished and no mongodump processes are running on client machine (windows).
It happened twice today, 1st occurence caused the P->S switch, since DB was almost unresponsive.
Any idea, what can cause this ?
Regards,
Matt,
Hello,
I face something similar. We deployed replica on AWS EC2 (in docker. Many of them, various instance types). We run daily backups.
Sometimes (once per week approx.) the CPU goes up during the mongodump and never comes back. The EC2 machine must be rebooted to become available again.
We did not face the same behavior on Vultr hosting (we have some replicas there as well)
I tried to call mongodump from a different computer to save recources on a database server. Did not help.
If anyone have any ideas, what could we do to make it stable again, if would be great.
Regards
Michal