Hi Philipp, On Fri, Jul 17, 2009 at 00:41, Philipp Kern<tr...@philkern.de> wrote: > Hi, > > On 2009-07-16, Sandro Tosi <mo...@debian.org> wrote: >> today ries (aka ftp-master) was down due to a scheduled maintenance activity. > > more or less scheduled, as already stated. > >> Now, scheduled means programmed, and suddenly this question comes to >> me: should the project be notified of such core activities? should we >> only relay on #debian-devel irc channel topic to know this? > > Interestingly that is even mirrored to a public location[1], especially as
I know, but I don't see irc as the best medium to notify of such activities. > downtimes at short notice have in the past been put there. While I agree > that they could have mailed d-i-a I, for myself, don't deem that necessary > for such a short outage. ehm, it was not "so short" :) it lasted some hours. Moreover, there were several other occasions in the past where a "formal" communication was not sent: for example, Fort Collins network issues some months ago. > DSA regularly update ries with new kernels and notice ries interactive > users sufficiently ahead, so I have nothing to complain in this regard. > (It's mostly on IRC, true, but for the current userbase that seems to be > sufficient because we're there when/while working.) I'm not saying DSA doesn't coordinate with key users before performing any action, just that an email to the project like "oh hi, we brought/will bring down <machine> @ HH:MM, we estimate X hours of downtime but as Murphy sits on the next desk, it may take longer" would also be appropriate. I don't think IRC is the right place for such communication, at least not the only one. >> ries is not a barely used machine, it's the one of our fundamental >> servers, and it went down without notification (to my knowledge, or at >> least to a wide audience). > > I tend to differ. Of course it's heavly used interactively by both > FTP and release team, but well, leaning back for half an hour also > does some good, right? > > There are (to my knowledge) three services served from that machine that > the public could access: 1) incoming, 2) the upload queue, 3) release > stuff like britney output. The upload queue was pointed somewhere else, > incoming might be critical for buildds but in general they cope if it > isn't reachable and the latter is partially put onto packages.qa.d.o. > >> Other times, when I asked on irc why a given machine when down I often >> have received a reply that sounded like "WTF you want? go and do >> something else and don't bother us", not really encouraging, but I >> moved along. > > I suppose you did get an indication why it was down, right? Maintenance > activity. So nothing is wrong per se, which is a good sign. People > are taking care of it. The only indication I have is inferred from the fact that the #-devel topic was changed by a DSA member using 'scheduled' to categorize the down. I think a clear statement of what's going on it's far better. >> Should we improve how we communicate in the project? Shouldn't there >> be more information on what's moving "behind the scenes"? It's just me >> that would like to know it? > > I find weasel's reports on what DSA did very interesting in the past. > Granted, they were internal notes put out into the public, but it > gives you some impression of "behind the scenes". and we must thank him for that report. Sadly I saw it just one time (plus the discussion that arose from it), but a continuous update for situation like ries is also valuable. Cheers, -- Sandro Tosi (aka morph, morpheus, matrixhasu) My website: http://matrixhasu.altervista.org/ Me at Debian: http://wiki.debian.org/SandroTosi -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
