Guest's picture

Hello,

Recently we have noticed that the automatic backups are seriously affecting our server performance. 

However, this morning the server became completely unresponsive and we had to reboot to get it back.

When we looked at the stats there were two processes consuming 100% of the CPU and disk io: tklbam and duplicity.

The backups now take hours instead of minutes, and we really don't have enough data to justify this amount of time.

Can anyone give me an idea about what's going on?

Thanks,

Dennis

Forum: 
Tags: 
Jeremy Davis's picture

As I asked in my email to you; Are you running a Micro server?

If so then, despite what your research suggests, it is probably not specifically the TKLBAM software (and Duplicity is the backend), but the combined server load while TKLBAM is running is going beyond the CPU limits set by Amazon for Micro servers. This will trigger CPU throttling. The throttled CPU will then in turn will make the server unresponsive (and make whatever is consuming the most CPU report 100% usage).

Issues with production Micro servers are not uncommon and unfortunately the only real fix is to go for a bigger server size. Essentially IMO Micro servers are only suitable for testing and development. Beyond a low traffic static HTML site you will eventually run into issues with a Micro server in production.

As a short term fix you may be able to limit the load that TKLBAM puts on your system (or even disable TKLBAM if you are willing to risk not backing up your data).

If OTOH you aren't running a Micro server then I wonder what might be going on...? Can you provide more details of your server as I have not experienced this issue and I have no other reports (other than Micro server users) of similar behaviour.

Add new comment