Anders Hedberg's picture

Hi, I often get an 504 error from Nginx.

The CPU graph shows about 30-40% activity, the Disk IO shows ~400MB and 70kB of net activity.

The Disk IO graphs seems flat on top, as they hit a limit and stays that way for almost an hour. Sometimes this extends to 1.5 hour, or even 2 hours, followed by a low for a short time of about ~0-5MB.

There are six cron-jobs that use wget to trigger PHP scripts from the local NginX. These scripts loads sensor values from another server, and then empties the local MySQL database and computes new median values for quick graphing on the webclients. (a better tool might be RRD Tool or Graphite, but this is our start).

It seems that the disk IO starts at every hour, so it is most likely the cron jobs that does this.

It is a micro instance on Amazon.

Assuming we hit a limit on disk IO, I would like to raise this limit by paying more, but I can't find any measures or other value on disk IO. The only thing I can find is something "Quadruple Extra Large" with SSD disks which costs $3100/hour. Which is far away from our current $0.02/hour and not really an option ;-/

Is there a disk IO limit? What is the Disk IO limit? Can it be raised? Is it fixed (i.e real IO performance on disk is maximised, so no help by upgrading to a mini instance?)

What is the 400MB we see in the Turnekey graph, is it IO transactions? Of what size? or MegaByte/second ?

Jeremy Davis's picture

But I do know that Micro instances can get throttled qutie easily with anything more than a low load. Personally I'd test a Small instance for a day or 2 and see how that goes...

Add new comment