is anyone missing any metadata? (patch tag just had another strange crash)

18 views
Skip to first unread message

Thomas Hartman

unread,
Jun 19, 2010, 7:00:57 PM6/19/10
to patch-t...@googlegroups.com, HA...@googlegroups.com
Hi, patch-tag went down and I just restarted manually.

However, I was unable to read state from the latest events file, which
means that the aplication state may be stale (a couple days old):

So, metadata for accounts / repos created in the last couple days
might be missing.

However, the darcs repos are still on the hard disk and should be
accessible via darcs push / pull, they just won't be listed.

I am investigating. Meanwhile, if anyone knows themselves to be
affected please let me know, including account names and other
relevant data (I can't think of any actually).

FWIW, this is the error I got on application start.

patchtag exited: Sat Jun 19 18:47:20 EDT 2010
starting patchtag: Sat Jun 19 18:47:20 EDT 2010
passed sanity check
starting happs server
time: Sat Jun 19 18:47:21 EDT 2010
patchtagserver: internal error: PAP object entered!
(GHC version 6.12.1 for i386_unknown_linux)
Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug

When I moved the latest checkpoint file out of the way it started normally.

Need somewhere to put your code? http://patch-tag.com
Want to build a webapp? http://happstack.com

Jeremy Shaw

unread,
Jun 19, 2010, 8:07:50 PM6/19/10
to ha...@googlegroups.com, patch-t...@googlegroups.com
Let me know if you figure anything out. Did you delete event files ?
Or just remove the checkpoint ?

If you only removed the checkpoint, then it should just replay all the
events and be up to date ?

The 0.6 released is primarily focused on improvements to happstack-
server (and happstack-ixset).
The 0.7 will be focused on happstack-state.

During 0.7, I would like to figure out why patch-tag crashes so much.
Because it shouldn't ;)

- jeremy

> --
> You received this message because you are subscribed to the Google
> Groups "HAppS" group.
> To post to this group, send email to ha...@googlegroups.com.
> To unsubscribe from this group, send email to happs+un...@googlegroups.com
> .
> For more options, visit this group at http://groups.google.com/group/happs?hl=en
> .
>

Thomas Hartman

unread,
Jun 20, 2010, 1:29:44 AM6/20/10
to patch-t...@googlegroups.com, ha...@googlegroups.com
I renamed the latest event file to eventnnnn.bak.

> "patch-tag-users" group.
> To post to this group, send email to patch-t...@googlegroups.com.


> To unsubscribe from this group, send email to

> patch-tag-use...@googlegroups.com.


> For more options, visit this group at

> http://groups.google.com/group/patch-tag-users?hl=en.
>
>

--

Jeremy Shaw

unread,
Jun 22, 2010, 4:47:02 PM6/22/10
to ha...@googlegroups.com, patch-t...@googlegroups.com
Just a a followup:

Investigating this issue, we discovered that some of his update
functions were calling the 'fail' method. This fail was not caught and
cause the whole server to go down when the event was replayed.

Clearly the handling of 'fail' is not correct. It's not obvious what
should happen when you call fail, but making your whole application
fail is certainly not what we want.

If you are *not* currently calling fail inside your query/update
functions, then you should be fine. So for now, just don't call fail.
In the happstack 0.7 we will reexamine this issue and make sure that
something sensible happens if you do call fail.

- jeremy


On Jun 19, 2010, at 6:00 PM, Thomas Hartman wrote:

Reply all
Reply to author
Forward
0 new messages