Network or server error on SAVE Bib record

[George Duimovich]

EG 2.1.1
OpenSRF 2.0.1

This record (even with with no modifications) will throw error: "Network or server failure. Please check your Internet connection to catalogue.nrcan.gc.ca and choose Retry Network" etc

Downloaded and posted marcxml here:
http://geoscan.ess.nrcan.gc.ca/evergreen/errors-8056134.xml

If I remove this URL, the record will save, so appears this incorrect URL was triggering error:

http://web.ebscohost.com/ehost/search/advanced?sid=38eb2872-9e28-4aae-9f30-e4fd2bb9ebdc%40sessionmgr14&vid=1&hid=19

Couldn't get at the JS console as my error window drops out of view and no way to scroll down to 'click confirm' etc. (even if I increase my screen resolution). Anybody care to test if this URL generates similar error for you too?

Perhaps another issue to address is to include scroll bar with "Network Failure" and other similar error windows as this error when off of the bottom of my screen.

[George Duimovich (george-duimovich)]

actual error message

[Dan Wells (dbw2)]

I strongly suspect this is due to an issue with the OpenSRF gateway. Similar problems were investigated last May, and while some progress was made on finding the cause, no solution was put forward at that time.

I have spent a fair amount of this morning getting into the OpenSRF code for the first time, and I think I might have the solution. I'll get it in a branch after lunch if testing bears it out.

Thanks,
Dan

[Dan Wells (dbw2)]

More explanation and proposed fix now at:

http://git.evergreen-ils.org/?p=working/OpenSRF.git;a=shortlog;h=refs/heads/user/dbwells/fix_printf_gateway_crashes

[Dan Wells (dbw2)]

This is really an OpenSRF bug, not an Evergreen bug, so I have linked it to that project as well.

[Dan Scott (denials)]

Awesome, thanks for a really important fix, Dan! Pushed to master and rel_2_0 of Opensrf.

[Jeff Godin (jgodin)]

Dan Wells-

I'm seeing an apparently similar issue with a record containing no % characters. When saving the bib in question, the staff client returns an error, and server logs do not acknowledge even receiving the API call to update the record.

Is there a reliable sign that this gateway crash has occurred?

-jeff

[Dan Wells (dbw2)]

Jeff,

At one point I did get a "%n in writable segment detected" error in my Apache error logs, but even that hasn't been consistent across servers/versions (and wouldn't make sense in a %-less case anyway). If you paste the problem record somewhere I could tell you if the fix here makes a difference. It may also help me (or somebody else) spot the real problem even if it isn't related to this fix.

Dan W.

[Lebbeous Fogle-Weekley (lebbeous)]

So this bug, it need not target Evergreen any longer? There's nothing to test or commit at this point, correct?

[Dan Wells (dbw2)]

That is correct, marking as invalid for Evergreen. Thanks.

[Dan Scott (denials)]

Jeff, if you ever reproduce this problem, please post your problem record for us in a new OpenSRF bug and refer back to this one?