Preventing a CT to take down the whole system

[greg]

Greetings

I have a CT with a MariaDB inside it; I noticed that when a very heavy sql-related load occurs, the whole host system becomes unresponsive for an extended period of time: I cannot even use an already established ssh connection.

Is there a way to "contain" a CT so I can at least keep my host ssh connection alive?

Thanks in advance

Regards

GR

 

[gfngfn256]

the whole host system becomes unresponsive for an extended period of time

Have you identified the root cause?
Is it Disk IO or Network IO or maybe CPU/RAM based? (Or a mixture).
Once you have it identified, you can probably do some things to isolate that load away from the host. For instance if it is NW load, use a different NIC for the container. IDK your HW at all - but I'm sure you have what to work with.

 

[greg]

Thanks for your answer. The root cause is probably a heavy sql query (such as a query involving several millions lines). During that time, the server just "disappear" (it doesn't even ping), and when it's back there's a gap in the RDD graphs of the Proxmox GUI. The memory graph is flat at 25% usage. 10 out of 12 cores are attributed to the CT. So I guess it could be disk IO?

 

[Noah0302]

Is the CT running on the same Disk, on which Proxmox is installed?
For Example local-zfs.

If thats the Case, maybe you could move the CT to another Storage Location and see if it is still happening.

 

[gfngfn256]

10 out of 12 cores

What happens if you only give the LXC 6 cores? IDK the workload of the LXC or the host cores capabilities - but if that stops a lockup/freeze on the host - you know which resource is causing this.

One other thought. The fact that no ping is answered during this period may show - but not definitely - that its a NW overload.