Thanks for your reply - I am reassured. My server is a slow machine with limited memory - not much better than a Raspberry Pi, but designed to run 24/7 - so it is not surprising that processes will conflict from time to time.
One warning in eight months is more than acceptable :-) Thanks again Allen C On 08/10/18 12:03, Ralf Hildebrandt wrote: > * Allen Coates <znab...@cidercounty.org.uk>: >> Yesterday I saw the following warning message in my logs:- >> >> 2018-10-06T14:11:19+01:00 geronimo postfix/postscreen[8194]: warning: >> psc_cache_update: btree:/var/lib/postfix/postscreen_cache update average >> delay is 151 ms > > Oct 2 02:01:40 mail-cbf postfix/postscreen[23257]: warning: > psc_cache_update: btree:/var/lib/postfix/postscreen_cache update average > delay is 343 ms > Oct 2 02:03:16 mail-cbf postfix/postscreen[23257]: warning: > psc_cache_update: btree:/var/lib/postfix/postscreen_cache update average > delay is 155 ms > Oct 3 18:34:07 mail-cbf postfix/postscreen[23257]: warning: > psc_cache_update: btree:/var/lib/postfix/postscreen_cache update average > delay is 137 ms > Oct 8 11:21:19 mail-cbf postfix/postscreen[65199]: warning: > psc_cache_update: btree:/var/lib/postfix/postscreen_cache update average > delay is 112 ms > > Quoting from > http://postfix.1071664.n5.nabble.com/psc-cache-update-td10059.html > > "It is a bit sluggish. The warning threshold is 100ms. It should not > take this long to insert one key/pair into the database. Perhaps your > system's disk is very busy, or you're on a VM slice, or your clock is > not stable. If this happens frequently you need to find out why." > > and > > "If this happens often, this means that postscreen cannot handle > more than 10 SMTP connections per second, or that your system clock > is jumping (as in: running inside a VM). > > I see the warning once a day on my lightly-loaded server with a > single 15kRPM disk under an ancient CPU; the timing suggests that > this happens while some cron job is doing house cleaning. > > I added this check because someone insisted on running postscreen > on top of an SQL database, and complained that postscreen performance > was erratic. After I added the warning he stopped complaining." >
