SqueakMap dead?

Wed Nov 24 21:24:30 UTC 2004

Thanks Ken, we are back on line now.

Some recent updates to SM seem to be missing, and I'm guessing that they
were probably in that image segment that was trying to save itself. I
don't know how the server side stuff works, but most likely if there
is a save every N hours or days, that would tell us how much stuff
might be missing. Can you tell from the time stamp of the previous
image segment what that gap might be? We could then follow up with a
message to the effect that "If you added or changed anything on SM
in the last N days, please re-enter your work."

For what it's worth, I entered a new release of TimeZoneDatabase
on SM right before it went down. So it is quite likely that I am
the person who triggered the problem. In case it helps with debugging,
here is what I did:

- logged in to my SM account (with Mozilla)
- went to my Squeak Map home page
- selected "edit releases" for TimeZoneDatabase
- changed a bunch of fields in the form to make a release 1.2.3
- clicked the "Save as new release" button

SM accepted the entry. Everything seemed to be fine at that time, but
shortly afterwards I (maybe 15 minutes or so, I'm not sure) I tried
updating SM from a Squeak image so I could verify that the new release
was working. At that point, the SM server had gone belly up.

Thanks again,
Dave

On Wed, Nov 24, 2004 at 02:42:20PM -0600, Ken Causey wrote:
> By the way, related to this:
> 
> I didn't put two and two together initially but it occurs to me that
> before I initially restarted it I noticed that the load on the server
> was very high.  I failed to check fully but I suspect the SM process had
> the CPU pegged.  The most common problem by far is a problem that
> appears to be related to saving the image segments.  I think gokr has
> speculated that it is a conflict which can occur if image segment saving
> is taking place and someone tries to change something at the same
> moment.  At any rate the fact that the load was so high indicates to me
> that the SM process may have been in an infinite loop.  Said loop would
> seem to be involved in image segment saving.  Whether there is a bad
> situation in general that can lead to an infinite loop when trying to
> save an imagesegment or if it is a result of data being modified while
> said is going on, I don't know.  Anyway, just wanted to state my
> thoughts on the matter.
> 
> Ken
> 
> On Wed, 2004-11-24 at 14:34 -0600, Ken Causey wrote:
> > OK, I didn't do all the checking I could have.  There was one of those
> > partial imagesegment output (.s) files so I've deleted it and restarted
> > squeakmap, let me know if there are still problems.
> > 
> > Ken
> > 
> > On Wed, 2004-11-24 at 13:50 -0500, David T. Lewis wrote:
> > > On Wed, Nov 24, 2004 at 11:15:40AM -0600, Ken Causey wrote:
> > > > I've restarted the Squeak Map Server.  Let me know if you have any
> > > > further problems.  I'm on #squeak for a while.
> > > > 
> > > > Ken
> > > > 
> > > > On Wed, 2004-11-24 at 16:47 +0100, Marcus Denker wrote:
> > > > > 
> > > > > Squeak Map seems to be not working...
> > > > > 
> > > 
> > > Hi Ken,
> > > Something is still out of whack. When doing "loadFullFrom: 'map1.squeakfoundation.org'"
> > > we get a checkpoint dohicky from the server with:
> > > 
> > >   HTTPSocket httpGet: 'http://map1.squeakfoundation.org/sm/loadgz?mapversion=2.0&checkpoint=1441'"
> > > 
> > > But the response is:
> > > 
> > > '<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
> > > <html><head>
> > > <title>500 Internal Server Error</title>
> > > </head><body>
> > > <h1>Internal Server Error</h1>
> > > Error: Checkpoint already exists!<p>
> > > <hr>
> > > <address>Comanche/6.2 (unix)  Server at 62.4.94.149 Port 8080</address>
> > > </body></html>'
> > > 
> > > 

> 