Problems After 4.6c Upgrade

[cswizard]

After upgrading to 4.6c, everything seemed to be working, data was being recomputed, then my NSLU2 became all but unresponsive. It's still serving up WDLive so the web server is running but it doesn't appear to be getting updates from my WMR200, the web interface returns "The connection to the server was reset while the page was loading." in Firefox, ssh sessions time out...

The NSLU2 has been shut down and restarted with the power switch but the results are the same.

Suggestions anyone?

[Vetinari]

My sheevaplug had nearly the same. The system load increased by 10x. Before the update it said nearly 80% system load now it is 800-1000%. After the reboot now it seems, that everything is working correctly.

[black23]

Hi,
I want to report same problem. After update on 4.6c my Sheeva after 6:00 goes very slowly, CPU usage is extremly high. After rebooot everything looks good. This problem repeating everyday from update.
Previous version was ok.
Boris, please help us.
Thanks

Config: Sheeva, MH 4.6c, WS 23XX

[cswizard]

Well, oddly enough, I left the NSLU2 running in whatever degraded mode it was in for a couple of days while I got up the interest in looking into the problem again. I shut it down for another reboot (the 4th since the upgrade), unplugged power, unplugged the memory stick, plugged everything back in, powered it up and...viola...it seems to be working fine. Weather data recomputed, System Load: 2.00, 1.21, 0.93 which doesn't seem too bad for the NSLU2, web interface and ssh responding, problem...solved.

Hope everyone else ends up good-to-go as well.

[admin]

4.6c has some problems with socket connections between processes. I am working hard on fixing that.

[cswizard]

Updating the situation...the success I had after rebooting with 4.6c was short-lived. After about 45 minutes the NSLU2 went unresponsive again and has done so 3 more times. When i saw 4.6d posted, I rebooted and managed to upgrade before the system went unresponsive again. Unfortunately, the problem still exists with similar if not identical symptoms.

[Rainman32]

I seem to be having similar problems with 4.6e.
Symptoms are first noticed when graphs stop producing, all of the rest of the uploads continue fine, but can no longer login to the system by web interface or ssh. reboot and everything is well again. I tried setting daily reboot a 00:00 but that didn't work, going to try setting for just before midnight to see if that helps.

Just figured out how to produce the system graphs so will report those back when it happens again.

[admin]

When you are on NSLU2 recomputation might interfere withWeb-Interface. May be web interface might react with a timeout when overall system load reaches a certain level.

Doing shutdown etc is not really helpful on this, as it will start a new recomputation about 5 minutes after restart.

Please try to login via ssh (user root, password meteohub) and see processes running and load via "top". This gives information what is happening. As Linux does not tend to hang, pressing reset, flipping power switch and other "cures" are just good to damage your USB stick's file system, so I would not recommend these.

[Rainman32]

I should have specified that this is on a Sheeva. unfortunately when this has happened, ssh locks up also.

I looked into it and if using the Meteohub scheduled reboot, it does not recompute does it? although given the symptom of the graphs going bad, maybe I do want it to recompute anyways.

[admin]

It is new to me that 4.6e does hang on Sheeva. Do you have additional software installed on it?

Scheduled reboots do not trigger a recomputation. Recomputation on Sheeva should not interfere with web interface in any way.

Does "dmesg" show SD card problems? What type of card are you using?

[Rainman32]

No additional software at all. Using a Transcend class 6 card (TS4GSDHC6). I have been unable to find any errors in dmesg, but then again can't access it when the unit is locked up.. I believe dmesg gets overwritten on reboot.

this is happening consistently on 2 different plugs with different station types even. One of them had an older load of 4.6b and a sandisk card originally. the sandisk would not even boot up again without reseating it, so went to the transcend which has been stable through power-cycle and reboots.
the other sheeva is a clean install with 4.6e and the transcend card from the start, it has been having identical issues.

[Rainman32]

A little more background on the two units. Both are uploading to Weather Underground, producing and uploading several graphs, processing testags for Saratoga Weather template along with the various WD-Live files. I have not been able to put my finger on anything specific, but it seems that the graphs are suspect as that is where the symptom first shows up... if the system is left in this state for some time, the subsequent graphs are missing data for that period.

[admin]

OK, that needs a closer look. Could you please open port 22 for me to inspection? Please email WAN-IP and login details etc to "info(at)meteohub.de".

[Rainman32]

info sent.

now here is something I never noticed before in dmesg:

[ 7120.610000] TCP(wget:13309): Application bug, race in MSG_PEEK.
[14747.900000] TCP(wget:25243): Application bug, race in MSG_PEEK.

[Rainman32]

Just to clarify, I have not made any changes to the Meteohub installation whatsoever. no system package updates or any other updates at all outside of Meteohub. I am having intermittent "lockup" issues on a Sheeva with 4.6e. Looking through earlier entries on this thread I see I am not the only one who has experienced this on Sheeva.

By lockup I mean that the web interface and ssh are no longer able to authenticate. Uploads continue with the exception of graphs are no longer produced. recording of data during this period ceases also. Another weird symptom is with Weather Underground, the uploads continue except for the Gusts data gets stuck at where it last was before the issue occurred. I believe that all of the symptoms can be attributed to to a disconnection of the file system.

I have just seen another post that suggest the same issue with an eBox 330.

My personal observation is that it does seem related to uploading of data. I think that somewhere there is a failure with the upload process that sends the whole system overboard and results in disconnecting the file system.


Is this happening to You? if so on which platform and any other pertinent details. Please chime in and help Boris get to the bottom of it!