tower@AI.MIT.EDU (Leonard H. Tower Jr.) (06/05/89)
1) Many of you will be seeing month old messages. Explanation appended. Just so you know. Please don't bother root@ai.mit.edu or postmaster@ai.mit.edu about it. Use info-gnu-request@prep.ai.mit.edu or gnu@prep.ai.mit.edu if you have something to say about the GNU mailing lists! 2) As long as you are all here: i) If you get a failure message from a mailer about a message from any mailing list, do NOT assume the message didn't make it to the list. Do NOT re-send it to the whole list. Such messages generally mean that only a few people out of hundreds or thousands on a list didn't get your message. Forward the failure to the -request address for the list, which is formed by appending "-request" to the list: LIST_NAME-request@LIST.SITE.DOMAIN ii) if you wish to subscribe, unsubscribe, or change your address ask the -request address for the list. Do NOT ask the whole list!!! thanx -len From: root@ai.mit.edu (Operator) Date: Sun, 4 Jun 89 15:03:45 EDT Subject: Unix System Woes The Unix machines and mail have been very unreliable. We wish to explain some of the recent problems, what has been done, and what needs to be done. The major topics are system down time and mail. (1) Pipe incident. The most recent cause of downtime was caused by a water main leak. To fix the leak, the water supply and the air conditioning had to be shut off on two separate nights. Power cycling machines frequently breaks them, and this time around we had lots of disk problems. This is a hardware failure. (2) Wheaties disk failure. Before the pipe incident, wheaties boot disk drive died. Wheaties is at the core of the lab's unix systems, and losing its boot disk caused many clients to be down and stopped mail delivery. We couldn't isolate the failure, so we had to reformat the disk. Reformatting disks is painful and takes 24 hours. This is a hardware failure. (3) Mail delivery. For the past several months there have been many problems with Unix mail delivery. These problems were caused by errors introduced into the mailer configuration several months ago when the Sun named life was made responsible for final mail delivery. Furthermore, at that time a new version of the mailer was installed that was incompatible with the release of Unix that we are running. The result was that if a message was not delivered immediately it was then queued, but usually never delivered. Some 5000 messages suffered this fate. This is a software problem. We've attempted to correct the mailer configuration (by bringing up a new operating system version) and believe that the messages are now being delivered more reliably. However, there are still problems with the mailer, though we believe they are not affecting delivery. Sun is currently investigating a bug report that we submitted. thanx -len