PerfGrid

We've completed the first two of three patches on all servers but one (kernel will be available tomorrow) - general downtime per server was about 5 minutes - one server (server5) did suffer from an extended downtime due to a race-condition within CloudLinux.

We run something called "CageFS" which isolate users from each other, this works by mounting common folders as virtual and hiding the rest of the system.

The race-condition itself resulted in the CageFS folders being created for the users before the folder where databases are stored in, got mounted.

As a result all connections via "localhost" (socket connections) would fail, and all connections via TCP to MySQL would succeed.

Our monitoring page didn't do this check, which means that discovering the bug in first took more time than it should, extending the downtime.

What we will do, is to implement a check that does both socket and TCP MySQL connections to ensure that the connections continue to work as they should.

We're sorry for the inconvenience caused by this extended downtime.

We want to say thank you to all our customers and all the support we got throughout 2017 and wish everyone a happy new year tomorrow.

As we do every year, we've posted a small review with some stats including uptime statistics for our servers.

You can read the review here:

2017 review 2017 statistics (per 16/11/2017): Served 5.79 billion requests (average of 183.60 per second) which is a 79.81% increase compared to 2016. Our peak traffic week was around Black Friday with 240.89 …