ccea3@rivm.UUCP (Adri Verhoef) (12/16/89)
In article <1594@bnlux0.bnl.gov> abrams@bnlux0.bnl.gov writes: >/usr/lib/news/expire -I -e 14 -E 30 -v 6 -n all >>logfile > The output from expire looks like: >expire: nohistory 0, rebuild 0, doarchive 0 : : >updating active file /usr/lib/news/active : : >looking at group alt.dev.null > maxart = 00000, minart = 00000 >looking at group comp.security.announce > maxart = 00 >======================================================================== >Expire thereupon hung, and ps ax shows expire doing- > >10141 ? I 0:00 sh -c /usr/lib/news/run.expire? >10147 ? I 0:00 sh /usr/lib/news/run.expire >10151 ? I N 55:09 rnews -U >13047 ? I N 0:02 rnews -S -p 891206084039ed >13049 ? I N 0:00 sh -c cat /usr/spool/news/.ina13047 | /usr/lib/news/checkgr >13051 ? I N 0:00 sh -c cat /usr/spool/news/.ina13047 | /usr/lib/news/checkgr >13055 ? I N 0:00 sh -c cat /usr/spool/news/.ina13047 | /usr/lib/news/checkgr >13178 ? R N 55:32 egrep -v |^alt[. ?]|^atl[. ?]|^aus[. ?]|^austin[. ?]|^ba[. This shows, that expire is not running anymore! I see there a `checkgroups'. Anyway, today I saw another `rnews -U' running into problems. My news version is B2.11.19 on a vax-630, Uniq System V.3.0. MINDISKFREE = 5000 I have been watching the behaviour of `rnews -U' after completion of `expire'. ** The scene: Someone posts a message, but the disk space is beneath 5000 blocks. Looking in $LIBDIR/log shows: Dec 15 14:49 local Out of space in /usr/spool/news. Dec 15 14:49 local queued <133@rivm18.UUCP> ng rivm.test subj ___ from ___ At 18:00, `expire' starts running. When the expiry is done, `rnews -U' starts running, but the disk space is still < MINDISKFREE. Anyway, `rnews -U' fires off `inews -S -h', logging: Dec 15 18:05 local Out of space in /usr/spool/news. Dec 15 18:05 local queued <133@rivm18.UUCP> ng rivm.test subj ___ from ___ Dec 15 18:05 local Out of space in /usr/spool/news. So `inews -S -h' stops. But `rnews -U' fires off a new `inews -S -h', logging: Dec 15 18:06 local Out of space in /usr/spool/news. Dec 15 18:06 local queued <133@rivm18.UUCP> ng rivm.test subj ___ from ___ Dec 15 18:06 local Out of space in /usr/spool/news. And again and again... So, the disk space will gradually get tighter and tighter, until something is done, but that maybe too late. UID PID PPID C STIME TTY TIME COMMAND news 798 512 14 18:14:08 ? 0:01 inews -S -h news 512 1 9 18:06:35 ? 0:06 rnews -U Now, if the disk space suddenly suffices, the problems will go away: Dec 15 18:14 local Out of space in /usr/spool/news. Dec 15 18:14 local queued <133@rivm18.UUCP> ng rivm.test subj ___ from ___ Dec 15 18:14 local posted <133@rivm18.UUCP> ng rivm.test subj ___ from ___ Dec 15 18:14 local <133@rivm18.UUCP> sent to rivm05 My `rnews -U' succeeded in writing 77 lines of text per minute in the logfile. That is more than 50K characters in ten minutes. ** Solving the problem. I think there are several options to be considered when disk space has dropped below MINDISKFREE and `rnews -U' is running: - Stopping immediately after detecting "Out of space"; - Stopping after a number of retries has been reached; - Sleeping for a while after detecting "Out of space"; - A combination of sleeping and retrying. -- Adri Verhoef (uunet!rivm!a3 || a3@rivm.uucp), site administrator of the National Institute for Public Health and Environmental Protection (RIVM).