swish.swi-prolog.org is 500ing. I’m at work, hopefully somebody will see this and deal.
Restarted with larger memory limits. 8Gb seems to be too small to support over 400 users, which mostly shows as the system not being able to create new threads and returning a 500 error. I raised the limit to 24Gb. We’ll see …
Next step cloudifying the server to handle the thousands of concurrent users?
not a bad ‘dog food’ exercise at all
That is the plan. Being partly state full by definition this isn’t really easy though. I’ve started work that allows swish instances to sync using UDP messages. No clue when that will see daylight.
SWISH has been very useful in trapping all sorts of errors in the past couple of years. Students are really good throwing erroneous programs against it and the server is sometimes under heavy load, stressing resource management.