# Upgrade without downtime? **URL:** \ **Category:** Deployment\ **Created:** [August 14, 2007, 10:49am UTC](https://www.ruby-forum.com/t/upgrade-without-downtime/112015 "2007-08-14T10:49:14Z")\ **Posts on this page:** 4\ **Page:** 1
**Author:** ![piet](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/a9adbd/32.png) [@piet](https://www.ruby-forum.com/u/piet)\ **Post date:** [August 14, 2007, 10:49am UTC](https://www.ruby-forum.com/t/upgrade-without-downtime/112015/1 "2007-08-14T10:49:14Z")
Hi List! I’ve been thinking for a while about how to upgrade a running rails installation without any downtime for the users. Let’s say you fixed some bugs without doing any changes to the db schema or anything that suddenly looks/works different and would spook the users. One path I’ve been thinking about: Let’s say you run 10 mongrels. -Upgrade your rails installation (svn update or scp or whatever you’re using) -Notify whatever is doing the load balancing that 5 mongrels will be stopped. (this is possible with HAProxy, which I’m most familiar with) -Stop 5 mongrels -Start those 5 mongrels again -Notify the LB that the first (new) 5 mongrels are running again and that the remaining (old) 5 mongrels will be stopped -Restart the remaining 5 mongrels -Notify the LB that all 10 mongrels are running again To have less of an impact on the load per mongrel you could restart the mongrels one at a time. But you get the basic picture. Any comments on this ? Is there a better way ? Will this cause problems ? How do other people do this ? Thanks in advance for your insights! Piet. ---
**Author:** ![piet](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/a9adbd/32.png) [@piet](https://www.ruby-forum.com/u/piet)\ **Post date:** [August 14, 2007, 7:45pm UTC](https://www.ruby-forum.com/t/upgrade-without-downtime/112015/2 "2007-08-14T19:45:45Z")
Hi- On Aug 14, 2007, at 1:47 AM, Piet H. wrote: > -Upgrade your rails installation (svn update or scp or whatever > > To have less of an impact on the load per mongrel you could restart > the mongrels one at a time. But you get the basic picture. > > Any comments on this ? Is there a better way ? Will this cause > problems ? > How do other people do this ? > > Thanks in advance for your insights! > > Piet. If you use monit to control your mongrels it does a rolling restart. By that I mean it gracefully restarts the mongrels one at a time. With nginx or something in front this will be seamless to your users. Cheers- – Ezra Z.-- Founder & Ruby Hacker – removed\_email\_address@domain.invalid – Engine Y., Serious Rails Hosting – (866) 518-YARD (9273) ---
**Author:** ![piet](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/a9adbd/32.png) [@piet](https://www.ruby-forum.com/u/piet)\ **Post date:** [August 14, 2007, 8:15pm UTC](https://www.ruby-forum.com/t/upgrade-without-downtime/112015/3 "2007-08-14T20:15:48Z")
On 8/14/07, Ezra Z. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote: > If you use monit to control your mongrels it does a rolling restart. > By that I mean it gracefully restarts the mongrels one at a time. > With nginx or something in front this will be seamless to your users. Do you mean ‘monit -g mongrels restart all’ is a rolling restart? – Alex ---
**Author:** ![piet](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/p/a9adbd/32.png) [@piet](https://www.ruby-forum.com/u/piet)\ **Post date:** [August 14, 2007, 8:23pm UTC](https://www.ruby-forum.com/t/upgrade-without-downtime/112015/4 "2007-08-14T20:23:49Z")
On Aug 14, 2007, at 11:15 AM, Alexey V. wrote: > On 8/14/07, Ezra Z. [removed\_email\_address@domain.invalid](mailto:removed_email_address@domain.invalid) wrote: > > > If you use monit to control your mongrels it does a rolling restart. > > By that I mean it gracefully restarts the mongrels one at a time. > > With nginx or something in front this will be seamless to your users. > > Do you mean ‘monit -g mongrels restart all’ is a rolling restart? Yes. The monit command will return immediately and it will roll thru the mongrels one at a time and restart them. Cheers- – Ezra Z. – Founder & Ruby Hacker – removed\_email\_address@domain.invalid – Engine Y., Serious Rails Hosting – (866) 518-YARD (9273)