mikel@codas.att.com (Mikel Manitius) (02/26/88)
Recently when upgrading to SVR3.1.2 on codas, I restored the two integral disks from backup tapes that were two days old. The other two (expansion) disks didn't need restoring since they weren't repartitioned. USENET was on one of those disks. Since the backups I restored from were a couple of days old (weekend), that meant that I had to bring the "history" file back up to sync with what was currently on disk by running a rebuild, no big deal. The "Gotcha!" came when I realized that postings being made by local users were being thrown away as duplicates! It quickley became evident that the "seq" (Local article number sequence file, used for the Article-ID) file was two days old. New postings would use an Article-ID that was already in the history file, and be discarded as duplicates, with no error messages other than what's put in the log file! Watch out for this gotcha! -- Mikel Manitius mikel@codas.att.com
eric@snark.UUCP (Eric S. Raymond) (02/29/88)
In article <3003@codas.att.com>, mikel@codas.att.com (Mikel Manitius) writes: >The "Gotcha!" came when I realized that postings being made by local users >were being thrown away as duplicates! It quickley became evident that the >"seq" (Local article number sequence file, used for the Article-ID) file >was two days old. > >New postings would use an Article-ID that was already in the history file, >and be discarded as duplicates, with no error messages other than what's >put in the log file! This problem disappears in 3.0, which generates its Internet IDs as <ttttt:ppp@xxxxx> where ttttt is a hex encoding of a timestamp, ppp is the posting process ID, and xxxxx is the usual domain name. No seq file need apply. One more good reason to convert as soon as the sources are posted :-). -- Eric S. Raymond (the mad mastermind of TMN-Netnews) UUCP: {{seismo,ihnp4,rutgers}!cbmvax,sdcrdcf!burdvax,vu-vlsi}!snark!eric Post: 22 South Warren Avenue, Malvern, PA 19355 Phone: (215)-296-5718
matt@oddjob.UChicago.EDU (Mr. nEtural) (03/01/88)
In article <3003@codas.att.com> mikel@codas.att.com (Mikel Manitius) writes:
) I restored two disks from backup tapes that were two days old. The other two
) disks didn't need restoring since they weren't repartitioned. USENET was on
) one of those disks.
) I had to bring the "history" file back up to sync with what was
) currently on disk by running a rebuild, no big deal.
)
) The "Gotcha!" came when I realized that ... the "seq" file was two days old.
) New postings would use an Article-ID that was already in the history file,
) and be discarded as duplicates. [edited]
I solved that and related problems here by sticking the lib directory
under the main news spool directory. Now it all goes or none if it
does. (If it all goes, I fetch it all from a neighboring campus site.)
Matt
hargen@pdn.UUCP (Bill Hargen) (03/03/88)
In article <14437@oddjob.UChicago.EDU> matt@oddjob.UChicago.EDU (Mr. nEtural) writes: >I solved that and related problems here by sticking the lib directory >under the main news spool directory. Now it all goes or none if it >does. (If it all goes, I fetch it all from a neighboring campus site.) There is also an advantage to having the spool and lib directories in separate file systems. This morning, for the second time in less than a month, we were bitten by the System V "out of i-nodes" bug. But since the lib directory was in a separate file system, the log file contained the message ID's of the articles that were trashed. I was able to do a little editing and create a sendme control message to send to the upstream site asking for the lost articles. This sure beats having to contact someone and ask them to resend all of the last 24 hours of news. Of course, in this case the info would still have been in the log file even if it was in the same file system, since it was just out of i-nodes. But if we had run out of free blocks and spool and lib were in the same file system, then there would have been no history of what was lost. -- Bill Hargen Paradyne Corporation {codas,usfvax2}!pdn!hargen Mail stop LF-207 Phone: (813) 530-8655 P.O. Box 2826 Largo, FL 34649-2826
henry@utzoo.uucp (Henry Spencer) (03/03/88)
> This problem disappears in 3.0, which generates its Internet IDs as > <ttttt:ppp@xxxxx> > where ttttt is a hex encoding of a timestamp, ppp is the posting process ID, > and xxxxx is the usual domain name. No seq file need apply. C News does the same thing, although in a slightly different format. We did it this way because the seq file was just too much of a headache, but in retrospect we consider it the right thing to do anyway. -- Those who do not understand Unix are | Henry Spencer @ U of Toronto Zoology condemned to reinvent it, poorly. | {allegra,ihnp4,decvax,utai}!utzoo!henry