LOL, well I am trying to do drill test and see how resilience of runit could be, this is one of the minor downfall.
And yes I did "killall -9 runsv" so simulate it have been kill by whatever reason. On Thu, Jun 23, 2016 at 9:37 AM, Colin Booth <cathe...@gmail.com> wrote: > On Jun 22, 2016 5:39 PM, "Thomas Lau" <tlau <t...@tetrioncapital.com>@ > <t...@tetrioncapital.com>tetrioncapital.com <t...@tetrioncapital.com>> > wrote: > > > > here is the run script: > > > > #!/bin/sh > > exec 2>&1 > > echo "*** Starting service ..." > > RUNASUSER="tlau" > > RUNASUID=$(getent passwd $RUNASUSER | cut -d: -f3) > > RUNASGROUPS=$(id -G $RUNASUSER | tr ' ' ':') > > exec chpst -u :$RUNASUID:$RUNASGROUPS /usr/bin/memcached -vvv -m 64 > > > Your runscript seems ok, I'd probably do the user lookup differently but > that's mostly a style thing. > > > > I just tested -P, doesn't help and I could kill runsv process, memcached > daemon still running. > > > > I know OOM might not kill it, just trying to simulate what happen, who > knows when I was working on a system at 3AM in the morning and accidentally > kill runsv? :) I want to find out how is runit fault tolerance level. > > Oh, it sounded like you were having real issues. Runsv (and the rest of the > runit suite) should stay functional well after the rest of the system has > been run into the ground. As long as you avoid issuing a "killall -9 runsv" > or something equally catastrophic you should be ok. Generally speaking, as > long as you use the sv command instead of direct signaling, you're good to > go. In my history of using supervision-heavy systems and getting paged > awake at 3 AM, screwing up the supervisor is low on my list of mistakes. > > Cheers! > -Colin > -- Thomas Lau Director of Infrastructure Tetrion Capital Limited Direct: +852-3976-8903 Mobile: +852-9323-9670 Address: Suite 2716, Two IFC, Central, Hong Kong