Fri Jan 29 17:52:33 PST 1999 — Update on the DSL slowdown. Covad has worked with Cisco, and at around 7am on January 27th, fixed the trouble. It affected all bay area clients who were using Covad services. -Dane
Tue Jan 26 14:55:24 PST 1999 — As it turns out, the Cisco 2514 routing our DSL circuits didn’t have the CPU horsepower to serve its interface queues, which degraded performance for DSL customers. We fixed that by moving the DSL circuits to our core router, which uses a ‘silicon switch engine’ to forward packets. The result: DSL customers should notice improved performance; additionally, they are one hop closer to the Internet. -Scott and Eli
Update (16:19) – while this network redesign has helped, customers are still experiencing poor performance. Further discussions with the telco have revealed that they are actually having a problem, which they’ve now acknowledged:
Last update : January 26, 1999 03:45:41 PM PST
Covad is currently experiencing a network problem in the San Francisco Bay area. Customers affected by the problem may experience packet loss. Covad Network Operations and Engineering support are actively working on the problem. We will post updates on this page as they are available.
We’ll keep you updated as we get new information. -Dane
Mon Jan 25 19:41:03 PST 1999 — The elimination of the requirement to use the .ppp suffix to request PPP access is complete. Per the MOTD entry of Jan 14, this only affects direct dialup shell users. Please see the previous entry for more details.
If you use PPP or telnet, you won’t notice any immediate change. However, users no longer need to add the .ppp to their login name when connecting via PPP, so this should make setting up with Sonic easier. If you’re currently using the .ppp, there’s no need to eliminate it however, both methods work. -Dane
Sat Jan 23 15:00:31 PST 1999 — Friday morning’s upgrade included an operating system upgrade for Storm — aka www.sonic.net — from Linux 2.0.35 to 2.2.0. This caused problems with Microsoft’s Frontpage 98 server extensions ‘for RedHat 3.0.3’. (RedHat 3.0.3 was a Linux 1.2.13 release dated March 2nd, 1996.)
The server extensions work (after much tweeking) with Linux 2.0.35. Unfortunately, they lose significant functionality with Linux 2.2.0.
We’ve reverted Storm to Linux 2.0.35, and we will open a dialogue with Microsoft about the compatibility issue. Sorry about the trouble; we didn’t anticipate this as a problem. Please visit news:sonic.help.www for more discussion on this and other FrontPage issues. (The thread is entitled ‘FP98 broken??’.) -Scott
Fri Jan 22 21:05:56 PST 1999 — During last night’s upgrades, we changed the way we translate routes from RIP to OSPF. This caused a problem: the new translation system prevented correct routing from our internal network to some dialup networks connected at remote POPs. While the affected networks still had Internet connectivity, many rely on our DNS servers, which they couldn’t reach. We reverted back to the old translation system this afternoon. We apologize for the trouble to anyone affected. -Scott
Fri Jan 22 08:23:08 PST 1999 — This morning’s server maintenance efforts came off successfully. All planned relocations and upgrades were executed.
The highlights: all major servers are now on centrally protected power, connected via 100 megabit switched ethernet, in force-ventilated, deathstar-like monolithic enclosed rack cabinets. Upgraded firmware and OS, another SCSI controller and two new disk shelves for the Network Appliance effectively double potential capacity; 12GB extra RAID storage was added immediately, mostly for user web data. Storm, the web server for www.sonic.net, received a kernel upgrade which appears to have fixed the long standing SMP problems; as such, the second CPU has been reactivated; CPU horsepower thus has increased up to 95%. The FTP and SSL servers had their physical RAM doubled to 128MB.
Service downtimes were approximately as planned: roughly 20 minutes each for news and FTP service; 5-10 minutes for multihomed and SSL web service, plus a total of 25 minutes for web, incoming mail and shell services during the NetApp upgrade.
All servers are online and appear to be functioning properly. -Dane, Scott, Brian, Devin, Ian, Asa
Wed Jan 20 20:43:12 PST 1999 — Friday morning, between the hours of midnight and 6am, we’ll be doing some major maintainence here at Sonic. Specifically, we’ll be moving all critical servers into a different rack with centralized protected power, adding capacity to and upgrading the operating system of the Network Appliance fileserver, and upgrading some miscellaneous server hardware. All services maintained by the Network Appliance will be affected. This includes web hosting, FTP, and mail. Dialup and network service will not be affected, though dialup and shell users will be temporarily unable to access their email. The interruption should last from 3am to approximately 3:30am. In addition, there will be a number of sporadic (maximum 30 minute) service interruptions between midinght and 6am while various servers are moved and/or upgraded. These interruptions will include: Usenet news, web, and FTP service. -Dane, Scott, Brian, Devin, Eric, Ian, Asa
Thu Jan 14 09:26:14 PST 1999 — On Monday January 25th, we’ll be changing the default login service from shell to PPP. This will affect current shell users who are using a dialup terminal program to reach the shell server. Starting now, please use ‘username.shell’ to log in and reach the shell server – after the 25th, logging in as ‘username’ will serve PPP by default.
This only affects customers who dial directly up to the shell server. SSH and telnet sessions are not affected. PPP users who use ‘username.ppp’ are not affected.
For ease of typing and memory, we’ve also enabled ‘username.sh’ and ‘username.unix’ to shell access. Please begin using ‘username.shell’ or one of these alternatives immediatly. -Dane
Thu Jan 7 19:50:10 PST 1999 — Buzz, the new mail server. Yesterday morning’s test of the new mail server (making it ‘prog-relay’) went well, except for a few problems: most importantly, email for email@example.com was either deferred or bounced for a few hours. (Sorry about that — and thank you, Eric, for figuring it out. 🙂
At 2am, January 8th, we will let it handle mail for _just_ the sonic.net domain. Email for other domain names won’t be affected.
Again, transitioning to the new server should be seamless, and should require no action by Sonic members. (With one exception: folks using complicated .forward files may want to examine ‘man dot-qmail’ on Bolt.) -Scott
Tue Jan 5 17:50:27 PST 1999 — New mail server testing. We have completed installation and configuration of ‘Buzz’, our new mail server. This system will eventually replace Sub, our old mail server. Transitioning to the new server should be seamless, and will require no action by Sonic members.
At 2am, January 6th, we will point the ‘prog-relay’ host alias at Buzz, which will only affect those members who use prog-relay. (If you don’t know what ‘prog-relay’ is, then you won’t be affected. 🙂 This will allow us to observe the system delivering real traffic — if there is any trouble that can’t be immediately corrected, we will revert to the old system. -Scott