Sun May 18, 2014
On 2014-05-17 we received an automated email from our elasticsearch provider found.no who informed us that our elasticsearch instance ran out of memory and was restarted. Unfortunately this email did not directly rang the alarm sirens on our end as we had this a couple of times before without causing any inconvenience.
We noticed the service was unavailable soon. We were able to quickly narrow it down to our elasticsearch index being corrupted. Since we don’t host our own elasticsearch but instead rely on the awesome service of found.no, we immediately reached out to their support.
Fortunately they were able to restore everything within minutes.
They also told us that our elasticsearch instance was badly in need of more memory and that the same problem will happen again if we don’t provide it with more memory.
What do we do to make sure this won’t happen again?
1.we immediately doubled the amount of memory of our elasticsearch instance
2.we will install email/sms warnings to get notified about outages quicker
3.we will provide a http://status.ninya.io site for everyone to quickly see the status of the service
How can you help us to make the service more reliable?
ninya.io is a free open source service. It’s currently backed with a small private hobby budget. If you like to become a sponsor please drop an email to firstname.lastname@example.org. Alternatively you might leave a gittip if that’s more your thing.