Month: July 2000

Sat Jul 8 15:12:26 PDT 2000 — We’re continuing to deploy broadband content for DSL customers, and we’ve added a new section to our home page for this material. We’re continuing to work with our content partners to bring additional streams online. The content partnerships and hardware deployments in place with Akamai, iBeam, Multicast ISP and Yahoo Broadcast have allowed us to bring quite a bit of new content in on satellite in into caching proxies. With three satellite dishes on the roof now, and two more in negotiations, Sonic.net is ramping up lots of new streaming media in the near future for broadband customers. -Dane

Sat Jul 8 15:04:43 PDT 2000 — To improve performance of our mail server array, we’ve turned off GCOS matching for incoming email. Previously, if an email came in for an address that was not a Sonic.net username, the mail servers would attempt to match that username against the GCOS, which contains the user’s full name. So, for every SPAM or misdirected email, the mail servers would search the entire passwd’s GCOS looking for that name. During times when we’re bounding piles of SPAM, this was a significant impact on the mail servers. This impact was causing rare cases where YP would fail, and we’d bounce some legitimate customer email back to the sender as ‘user unknown’.

Our concern was that turning off the GCOS matching would break first.last@sonic.net delivery addresses which customers might have been using. Users could previously receive email as John.Doe@sonic.net for example. An analysis of the few million emails we’ve processed here in the last 10 days was done by Nathan and shows that no one uses this feature, so we’ve turned it off to eliminate the impact on the servers. If you’d like an alias for first.last@sonic.net, or any other email alias, just email support@sonic.net and we’d be happy to set one up for you. -Dane, Scott and Nathan.

Fri Jul 7 01:15:05 PDT 2000 — After 12 hours of non-stop work the news server is back online. We apologize for the extended downtime. Many thanks to our friend Ross for his assistance with Solaris. -Scott, Eli Kelsey and Dane

Thu Jul 6 12:15:48 PDT 2000 — We’re taking the News Server offline for 10 minutes to upgrade its BIOS. This hopefully will fix the problems we’ve been having integrating our new RAID array. The new RAID array will hold ‘overview spools,’ which will substantially speed up the time to open a new newsgroup. -Kelsey, Eli, and Scott

Thu Jul 6 23:32:40 PDT 2000 — After much bad luck with system corruption and attempted upgrades gone wrong, we’re in the process of a new installation onto a fresh hard disk for our Usenet news server. Once that’s complete, the /opt filesystem with the Typhoon news software, databases and spools will be mounted to bring news back online.

Meanwhile, we’ve of course got our redundant backup news system at supernews.sonic.net that you can use. It doesn’t have the sonic.* local groups, but you’ll find it to be quite complete and responsive.

Scott, Eli and Kelsey are pulling the night watch to bring this system back online, and I’ll be covering the early AM shift while they get some much needed rest Friday. -Dane

Thu Jul 6 19:35:13 PDT 2000 — After many Solaris nightmares, we can confirm that we most definitely love Linux much more than Solaris x86. We’re escalating with Sun’s engineering group to get additional assistance, and we hope to wrap this up as soon as possible. Basicly, the BIOS upgrade caused the PCI bus to enumerate the devices differently, and this confused the operating system.

Thu Jul 6 13:43:42 PDT 2000 — The News Server BIOS upgrade didn’t go as planned – the new BIOS appears to have changed the IO port and/or IRQ for the SCSI controller, which Solaris doesn’t like. We are currently scrambling to get it back online ASAP. -Kelsey, Eli, and Scott

Wed Jul 5 09:10:30 PDT 2000 — Our 1003 dialup group was returning all circuits are busy messages this morning. We discovered a PRI that needed rebooted. After the reboot everything is fine. Some people may have been disconnected when this happened. -Steve

Wed Jul 5 16:12:50 PDT 2000 — The News Server is going offline for more work on its RAID array. We’ve encountered an incompatibility with the existing SCSI controller and are going to swap it out for a replacement. Downtime shouldn’t exceed 20 minutes. We are sorry for the inconvenience the news server maintenance is causing but we are working on it in the fast track to resolve it’s performance issues. -Kelsey and Eli

Wed Jul 5 14:31:24 PDT 2000 — The News Server is offline for the integration of its new RAID array. Total downtime should not exceed 20 minutes. – Eli and Kelsey