Running on windows server 2003.

Mysteriously from time to time i find ELOG has crashed and i must restart it.

Event log shows:

This has occurred twice so far in the month or less elog has been running.  The platform is stable and i have no other issues or negative events besides incorrect ASP calls here and there.  Thanks.

Faulting application elogd.exe, version 0.0.0.0, faulting module elogd.exe, version 0.0.0.0, fault address 0x0004b9c3.

For more information, see Help and Support Center at http://go.microsoft.com/fwlink/events.asp.

 Unfortunately this is not enough information for me to conclude anything. You need to be able to reproduce the crash, then tell me exactly how you did it, so that I can reproduce it. Then I have a chance to fix it.

I have two logbooks running, one for current accounts and one for non-active ones. The idea is to move an entry from one logbook to the other once an account becomes non-active. Both logbooks share some attributes (e.g. the name of the account, its creation date, ...), but both also have some uniqe attributes.

While I was playing with this set-up, I ran into a problem:

1. Move an existing entry from current to non-active.
   • "Type" and "Status" have options. "Created" and "Archived" are in date format.
   • "Created" and "Type" are supposed to be locked in non-active. This works.
   • "Archived" and "Status" are empty, as they do not exist in current. This also works.
   • Depending on the "Type", only certain of the additional attributes are shown on edit. This works as well.
   • I would also like "Status" to have some influence on how certain other attributes are preset and/or shown. This is where the problem starts...
2. Edit the entry in current.
   • As expected, "Created" and "Type" are shown as locked.
   • As soon as I change the selector for "State", the whole edit mask is redrawn - and the date shown in "Created" is deleted.

I've been experimenting with this quite a while and as far as can see the following applies:

• Only date format attributes are affected
• Only locked attributes seem to be deleted completely
• If I have a date attribute that is not locked and I set it partially in the edit mask (e.g. only select the month or enter a year), it is also deleted
• If I have a date attribute that is not locked and I set it completely in the edit mask (i.e. day, month and year), it is not deleted

During the experiments I came up with a stripped down version of the non-active logbook configuration, which does show the bug. It only contains Type, Created, Status, Archived. To make an entry to experiment, you'll have to comment out the "Locked attributes" line, make a new entry (with "Created" date), the uncomment that line.

I hope my description isn't too confusing - please don't hesitate to ask if anything is unclear. I'd be much obliged if you saw a chance to fix this at some point...

Best regards,

Thomas

Hi

I am experiencing trouble with Elog crashing. I think it crashes during input operations (new entries are being added).

Are there any preferred methods to troubleshoot this ? Any log files I could look at ?

For now, I let "monit" supervise the process and restart it if it crashes. But I would like to get to the root cause of the problem.

Soren Poulsen

CERN

Dear Soren,

sorry for the trouble, but it's very hard to diagnose this remotely. This forum here runs fine for many months, so in principle Elog should be stable (at least the current version 2.7.5). If you have an older version, please upgrade. If you have a way to reproduce the problem reliably, you can send me your config file and explain step-by-step how you make Elog crash. Then I can try to reproduce it here. Otherwise the only chance I see is to run elog from inside the debugger, and once it crashes record the stack trace and send it to me. If you don't know how to use the gnu debugger, I'm sure you find someone at CERN who knows. 

Best regards,

   Stefan

Dear Stefan,

Thanks for your reply.

I started running elog in February and it never failed. Then it started failing regularly towards the end of September. There were no system changes until then, except the daily automatic Yum updates (SLC4). Then I upgraded to the latest version (tar ball of 2.7.5). Then yesterday it crashed again. I saw that it crashed around the time when a user was doing something - inputting new data. I can monitor when it crashes and correlate it with user activity. But it is not easy to reproduce since I don't know exactly what the user is typing.

It would be necessary to record the user input forms and then replay them against a known server state. But that is not so easy.

I will think about doing something else - maybe running inside a debugger as you suggest.

Soren

elogd is crashing with a segv when I try to create a new logbook by clicking the "Create new logbook" button
with an existing logbook selected in the dropdown list as a template. I've attached my config file for
reference. As a specific example, I'm logged in as the admin user, viewing logbook "1248", which has no entries,
and I click "Config" -> "Change config file" -> "Create new logbook", Select "1248" from dropdown list, enter
"1250" for the new logbook name, and click "Create new logbook".

The server crashes, and my web browser has the following as the (incorrect) URL it's trying to load:
(Incorrect because the "/ATLAS" shouldn't be there.)

https://localhost:8080/ATLAS/1250/?cmd=Config

Nevertheless, the config file and logbook directory are properly modified for the new logbook, and on restarting
the server, everything is fine.

Oh, just noticed that if I manually enter the URL:

https://localhost:8080/ATLAS/1250/

or actually, any text in place of the "1250", it also crashes the elogd process. So, perhaps there are two bugs?

(1) Incorrect URL being given to the web client after logbook creation
(2) Requesting any incorrect URL of this form crashes elogd

I just ran into the following bug:

I have a logbook where entries have several attributes, among which several dates. All of these are set to "Type <attr> = date". If I use the "Select" action, tag several entries and subsequently chose "Edit", the values of all date attributes are wiped. All other attributes are kept at their original values, unless changed explicitly. For the date entries, the date choosers are shown (as when editing a single entry), but all set to blank.

Editing single entries works fine.