Roller NetworkRoller Network has used Global Crossing as a transit provider for our bandwidth mix over the last year. Earlier this year Global Crossing was purchased by Level 3 and we’ve started to receive notices that as of this month our Global Crossing account will now be a Level 3 account. We don’t expect any changes that will impact us operationally (although we’ll be watching closely), and we will be referring to Level 3 as one of our transit providers from this point forward.
This weekend we’re going to be building a new server for the internal system that runs Asterisk for our office and the Request Tracker ticketing system that handles email for role accounts (like support and billing). Although the current server is still faithfully running with its original Athlon with 786MB of RAM, it’s starting to show its age (RT runs exceptionally slow) and the time has come to give it a forklift upgrade. As such, messages set to support, billing, and sales will be monitored manually (we’re using Mail Mirror on our domain) but they won’t get the usual auto response until after the new server is back online. Our office will also be unavailable by phone during the upgrade, although we rarely receive any calls on the weekends.
The new server we’re building for this upgrade will be an Intel Xeon X3430 with 8GB of DDR3 RAM on a Supermicro X8SIL-F motherboard in a 4U case instead of a 1U since we need to install multiple cards for Asterisk and the drive array controller.
UPDATE: We’ve finished restoring all of the major services and we’re working on the rest of the clean up.
We tracked down and fixed an intermittent bug in our Asterisk dialplan where a call could be dropped when using one of the group dial options (i.e. “press 2 for sales”). The issue was that one of our internal IAX peers was answering and dropping the call if it couldn’t establish a SIP channel rather than sending unavailable/congestion. We’ve altered our dialplan to send congestion explicitly when warranted.
This issue did not affect the 24-hour pager option since it always goes directly to a special voicemail box rather than a group dial, nor were direct dial extensions affected.
We have an unplanned Eaton service visit today at 17:00 UTC-8. There was an odd behavior with the UPS the resulted in a utility feed side breaker trip, so we’re having them come inspect the equipment, download logs, etc. The result was that the affected UPS ran on battery for about 5 minutes while we investigated and reset the utility feed breaker; there was no impact to our facility operations.
This event is a good example of where batteries have an advantage over a flywheel system. Since the input breaker tripped, the UPS was relying on its DC supply (whether it’s batteries or flywheels) to support the protected load. A flywheel would not have lasted the 5 minutes it took us to check everything. Flywheels, of course, have a lower long term maintenance cost since batteries are consumable items that need to be replaced on a regular basis. The tradeoff is that flywheels do not have the run time endurance of batteries.
UPDATE: No anomalies were found in the UPS performance monitors.
As promised, we’re going to be upgrading our PowerDNS authoritative servers from 2.9 to 3.0 in order to address the previously mentioned long TXT records and AXFR support limitations. We will perform the upgrade on Saturday, August 13 starting at 10:00 UTC-8. We expect the upgrade to be transparent.
At that time long TXT record support will go live (they’re already supported in the ACC, just not publishable) and we’ll start working on the front end to manage AXFR. We appreciate your patience.