I created a Digital Ocean VM that should be sufficient for our needs (if not, we can resize it or add others to handle greater load and/or bandwith/storage needs).
So far, I’ve installed the buildbot server. It is now available at http://buildbot.octave.org:8010. I believe @siko1056 will be able to help fix the web server configuration to eliminate the :8010 part of the URL and to make https work instead of just http, as we did with the Fosshost setup.
Unless access to the Fosshost system is restored, we have lost the history as that apparently wasn’t something we were backing up. Losing it would be unfortunate, but not fatal. Also, if we start collecting new build info before restoring the old database, I’m not sure how to merge them, but maybe we could get help with that.
The buildbot server is running and knows about the collection of workers. It looks like the workers need to be restarted to resume building. I did that for my worker systems and some builds are happening now. Windows builds won’t work again until we restore the mxe-octave mercural archive at hg.octave.org.
There’s a few things out there that provide free hosting for static websites only but do have limits on the traffic. The one I’m most familiar with is netlify which has 100GB per month free (or $19 per month for 1TB) and it’s been pretty stable for me (you don’t have access to a machine, they build the website and put it live when the repository changes, but I would count that as a plus). Anyway, that could solve the problem of the static sites (www and docs) but would at least be two of them solved if we know how much traffic goes to them.
webmasters@gnu.org keeps receiving messages of Octave users who can’t access the wiki, the download area, the docs, etc. We try to help them with links to this thread, to the Wayback Machine or to ftp.gnu.org. But they usually need to wait several hours for an answer. I think it would make their life easier if you disabled the rewrite directives in .htaccess that bypass the Octave pages on gnu.org. If needed, you could even add a prominent warning on the index page saying that the real website is down and will be back soon. (This is just an idea, not a request.)
We apologise for taking so long to fix it. We’re doing everything we can to fix it as soon as possible, but we’re dependent on the sponsors’ support team to fix it
“hg pull” does not pull the latest changes from hg.octave.org. Works fine if I use savannah directly.
$ hg -v pull
pulling from https://www.octave.org/hg/octave
real URL is https://hg.octave.org/octave
searching for changes
no changes found
(sent 4 HTTP requests and 2573 bytes; received 1666 bytes in responses)
$ hg id
0de399ce753c tip @
But:
$ hg -v pull
pulling from https://hg.savannah.gnu.org/hgweb/octave/
searching for changes
no changes found
(sent 4 HTTP requests and 1416 bytes; received 608 bytes in responses)
$ hg id
bbd028c2c233 tip @
Did we communicate to Fosshost that our move to alternative hosting is not just temporary (while their service is down) but that we probably won’t come back once (or if) their hosting is restored?
@mmuetzel: Not yet. I was intending do a couple of things first:
Create a maintainers thread asking what is to be done with the Fosshost space even if it comes back up. Choices are to surrender it, or to use the space for any non-critical use anyone might have. Right now, the host is still down and that discussion would be entirely hypothetical, so I haven’t created that yet.
Inform Fosshost of whatever we converge to in that thread.
I had tagged the Fosshost volunteers @peer and @sr229 early in this thread so they’d have a chance to be aware of this thread anyway.
Also, before we abandon the fosshost system, I’d like to recover the buildbot state.sqlite database from it and see whether it is possible to merge it with the new buildbot history that we are generating on the digital ocean system.
XXXXX — Yesterday at 7:20 AM
Update: We got reply from the dc, but as Thomas hasn’t given anyone access to their portal, we cannot do anything to fix it. I have even tried calling Thomas w/o any luck.
I would urge people to not wait on their vms and deem them lost and look for hosting elsewhere.
…deleted…
XXXXX — Yesterday at 7:54 AM
People who are dealing with our Chicago outage: please contact our ceo @Thomas (eth01) as only he has the ability to fix the outage. The volunteer team has done as much as we can and we simply need access to a portal Thomas ignores every request for access to. Sorry for the inconvenience.
October 31, 2022
I assume “DC” means “FDCServers.NET” and “Thomas” is “Thomas Markey” (CEO and director of fosshost).
Well, that’s unfortunate. I’ll continue to check on the host status and the VM we had there – if it ever comes back up again in a usable state, I can try to salvage the buildbot history that @jwe referred to, but other than that we’ve restored from backup and moved on.
