jrb@petro.UUCP (02/10/87)
We have had problems with the cron daemon summarily taking unexplained vacations. Local authorities are stumped, the cunning crontab crook comes creeping under curious circumstances, usually when the machine has been up and running for some period of time. Although the existing cron daemon can be killed and a new one restarted but I personally prefer to re-boot the computer to solve the problem. "Death of a cron run" occurs infrequently, but it is terribly frustrating. If anyone has solved this problem or at least experienced it, please let me know.
lyndon@ncc.UUCP (02/14/87)
In article <124@petro.UUCP> jrb@petro.UUCP (Jon Boede) writes: >We have had problems with the cron daemon summarily taking unexplained >vacations. Local authorities are stumped, the cunning crontab crook comes >creeping under curious circumstances, usually when the machine has been up >and running for some period of time. > [...] >"Death of a cron run" occurs infrequently, >but it is terribly frustrating. I have seen this happen under CTIX as well. It happened on a very sporadic basis. No core dumps were found. The only evidence was a error return in the accounting logs when cron tried to fork at startup. We don't have source, however I grabbed a copy of the binaries on the failing machine and brought them across to our (identical) system. A byte for byte comparison showed the files to be identical. I then shipped our /etc/cron to the other system and installed it. No problems noted for over three weeks. Hmmm... Installed suspect version on our system. This time it lasted three days before failing... Again a byte comparison was done. Again they matched. We eventually reinstalled everything from the distribution tape, and haven't run into the problem since. It's still one of the great unsolved mysteries though... -- Lyndon Nerenberg - Nexus Computing Corp. - lyndon@ncc.UUCP UUCP: {ihnp4,ubc-vision,vax135,watmath}!alberta!ncc!lyndon
abe@j.cc.purdue.edu.UUCP (02/19/87)
In article <1337@ncc.UUCP> lyndon@ncc.UUCP (Lyndon Nerenberg) writes: >I have seen this happen under CTIX as well. It happened on a very >sporadic basis. No core dumps were found. The only evidence was a >error return in the accounting logs when cron tried to fork at startup. One thing to check here: some Bourne shells run out of table space for &'d processes when executing /etc/rc and fail to start daemons in an unpredictable fashion. Inserting a few strategic "wait" calls in rc can sometimes help.
vrs@omssw2.UUCP (03/02/87)
I've seen this caused by a marginally too small stack size. Try upping the stack size to something reasonable with fixhdr if your cron is built with 0x100 bytes of stack or so. The default of 0x1000 should work much more nicely.