G. Pape
runit

runit - service dependencies

runit’s service supervision resolves dependencies automatically for many or even most service daemons, those that exit if another service they need to function properly is not or no longer available. The runsv supervisor will then start them again, until everything is in shape.

For service daemons with different behavior, or to arrange services more structurally, the sv program can help. A few examples:

For the service daemon floyd, that needs the pinkd service up and running, add sv start pinkd to floyd’s run script, right after set -e.

 #!/bin/sh
 set -e
 sv start pinkd
 exec floyd

When the pinkd service daemon crashes or is restarted for other reasons and the floyd daemon is unable to handle this, add sv hup floyd (or whatever the floyd daemon understands) to pinkd’s finish script to force its reload or restart.

 #!/bin/sh
 sv hup floyd

If the pinkd service shall be stopped when the floyd service is stopped gracefully, add test "$1" != 0 || sv down pinkd to floyd’s finish script.

 #!/bin/sh
 test "$1" != 0 || sv down pinkd

Or vice versa.

Explore the sv program for more options.

For service daemons that do not make their service available immediately after starting, a check script can be added. The sv program will use this check script to better evaluate whether a service is fully started and available, for example:

For a mail service listening on localhost:25, this check script could be used

 #!/bin/sh
 set -e
 exec nc -z localhost 25

Gerrit Pape <pape@smarden.org>