From: Hal S. <ha...@va...> - 2007-07-19 17:52:35
|
We have been testing a system with several Erlang nodes running yaws for recovery from various failures. We do such rude things to the system as rebooting servers while they are running and halting nodes which are then restarted from /etc/inittab. Yaws is a huge win, BTW - our toolchain now generates standard status and control pages for all Erlang apps. In some 50 or so restarts of this nature, we have seen two instances where a node refuses to start, exiting with the message =ERROR REPORT==== 16-Jul-2007::17:40:26 === There is already a yaws system running with the same id <"some_id"> on this computer, set another id in the yaws conf file We checked when this happens, and there is no other yaws instance running. Work-around is to remove the .../var/run/yaws/ctl-<id> file and restart. Probably what is happening is that something else is able to start listening on the previous yaws control port since the ctl- file was written and the old yaws instance died. This has a consequence for high-availability systems in that occasionally unattended restart fails. This is with Erlang R11B-5 and yaws-1.68. |