I get the same 503 erros when I run a gracefull restart from the litespeed web admin pannel.
I haven't got restarts every 20 minutes like in that topic.
I can't have 503 errors every time a client adds a subdomain…
The problem here is not what is causing the gracefull restart but why PHP becomes unavailable for almost 10 seconds just after the gracefull restart.
I don't think a gracefull restart shouldn't affect PHP processes… I've got the impression that litespeed restarts then asks all PHP children to restart and during the time it takes PHP to restart all children Litespeed gives out 503 errors because it can't connect to PHP.
I've only got 5 production sites on this server, my guess is it could take longer to restart all PHP processes when there are alot more children…
Could this be a setting or a bug with LSAPI PHP Suexec in deamon mode ?
Litespeed is sending a call to a PHP process that doesn't exist anymore because it has been restarted.
Where would I look to find out if it's a bad setting ?