Arc Forumnew | comments | leaders | submitlogin
4 points by albertb 6129 days ago | link | parent

I have more challenge for you.

Do the same thing, but make it survive a reboot of any server in the cluster.



2 points by s3graham 6129 days ago | link

I'm not sure I'd use db-transactional-serializable-migratable closures that often even if I had them (assuming the overhead wouldn't be worth the occasional "session timeout" message the user would see). Perhaps in some domains though?

-----

2 points by albertb 6128 days ago | link

You can't? You lose then. Kinda different those are, business and helloworld, aren't they?

-----

1 point by mst 6127 days ago | link

Looking at clients' deployed applications, the web boxes tend to average over a year's uptime so I can't say it seems like it'd matter that much.

Provided you've got runtime code loading and a useful QA process, I don't really see why you'd need to care - MySQL's lack of transactional DDL causes far more downtime than lack of serialisation would were we using a continuation-based system.

-----

1 point by albertb 6127 days ago | link

"web boxes tend to average over a year's uptime so I can't say it seems like it'd matter that much."

Unless there is a five or six nines SLA.

"Provided you've got runtime code loading and a useful QA process,"

Hardware might crash as well.

"MySQL's lack of"

That one is really good.

-----