Month: November 2015

One of the two pairs of storage clusters we use for email storage had a cluster takeover event at 11:09 this AM.  One of the heads lost communication with all of its disks and its partner successfully took over all of its services without any interruption.  However, while tracing the failure of the first head to a failed FCAL optics package – which was replaced – the cluster interconnect adapter in the second head locked up and triggered a panic.  This panic lead to a brief interruption if POP/IMAP services at approximately 12:00.  The second filer rebooted successfully still in partner takeover.

Unfortunately, since the second filer is still in takeover mode and can’t see the cluster interconnect adapter, the most conservative resolution requires that we halt the second filer and replace the failed adapter.  We’ve tentatively scheduled this for tomorrow after midnight provided that we receive the replacement adapter from our vendor in time.  POP/IMAP services will be offline for the duration of the maintenance which should take less than an hour to complete.

-Kelsey and William

Update 01:00: The cluster interconnect adapter has been replaced and all services have been fully restored.

-Kelsey and William

Tonight, November 25, at 1:00am, we will be performing a maintenance reload of core network equipment serving our Santa Rosa data center. No service interruption is expected, however in the worst case scenario customers may experience a brief period of routing instability towards colocation customers and Sonic services such as email and DNS.

-Tomoc

Ops has observed and fixed a problem with ftp.sonic.net that would have caused problems for users from 5:45pm to 8:50pm this evening. A networking issue between our VM cluster and the storage backend was to blame, and we are doing what we can to prevent an outage of this nature in the future.

– SOC

This evening, November 18, starting around 5:00pm, Amazon began blackholing traffic towards the Sonic network. We attempted to route through a different upstream provider, but it appears the routing issue was too deep into the Amazon network. We believe they have fixed the routing issue, and all traffic has been restored as of 5:25pm.

Update: This issue appears to have re-surfaced. We are reaching out to Amazon to determine the cause of the outage and do everything we can to ensure it does not happen again.

-Tomoc and the NOC