Changes between Initial Version and Version 1 of Ticket #67594, comment 3


Ignore:
Timestamp:
Jun 8, 2023, 4:43:30 PM (11 months ago)
Author:
ryandesign (Ryan Carsten Schmidt)
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #67594, comment 3

    initial v1  
    33If you're suggesting routing all build.macports.org web traffic through another server on a different internet connection so that it could provide a message when the real server is unreachable, which server would we use for that?
    44
    5 Conceivably we could use Braeburn, the server that powers www.macports.org and trac.macports.org. But I'm not sure if it has the capacity to accommodate that additional traffic. And it is in Germany and the buildmaster is in Texas. I'm not sure if introducing additional network delay to all requests is best, since people already complain that it responds slowly (which is both because of limited upstream network bandwidth and the amount of time it takes buildbot 0.8 to respond to some types of requests). It also introduces an additional point of failure. I usually keep a close eye on the buildmaster and know when it is down and what I need to do to get it back online, but that changes when additional servers outside of my control are added in.
     5Conceivably we could use Braeburn, the server that powers www.macports.org and trac.macports.org. But I'm not sure if it has the capacity to accommodate that additional traffic. And it is in Germany and the buildmaster is in Texas. I'm not sure if introducing additional network delay to all requests is best, since people already complain that it responds slowly (which is both because of limited upstream network bandwidth and the amount of time it takes buildbot 0.8 to respond to some types of requests). It also introduces an additional point of failure. I usually keep a close eye on the buildbot and know when it is down and what I need to do to get it back online, but that changes when additional servers outside of my control are added in.
    66
    77When I shut the servers down in anticipation of a thunderstorm and possible power outage, I leave the internet connection and router on, and the router could conceivably act as reverse proxy to provide a message when the backend is down. It might also be nice to let the router handle TLS and certificates. I did attempt to set that up about a year ago but didn't complete it.