Ok,

 

4th full system lock up on the latest RC.

 

Diags attached

Syslog attached (same one as in diag.zip)

 

I had Fix Common Problems running since I made my last post... The last diag.zip I found was at 9:10am GMT 14th Sept... Which means the lockup happened sometime between then and 9:30am GMT.

 

Hopefully this helps! Happening too frequently haha  Fingers crossed it's obvious what's wrong

 

 

raptor-diagnostics-20170914-0910.zip

syslog.txt

Updated to this beta from 6.3.5 (hoping to solve my crashing issues, but that's a different thread) and my SSD Cache became unmountable. Drive is mounted with XFS. Similar to the issue pointed out here https://forums.lime-technology.com/topic/59846-640-rc8q-cache-drive-unmountable-invalid-partition/

 

Went back to 6.3.5 and everything working fine again

 

I just upgraded from rc7a and upon reboot, the array was not started.  I checked and saw disc 3 was empty.  I wasn't sure which disc should go there at first glance, so I just rebooted, and this time, it started the array. YAY!

 

However, I went to the docker tab and it was totally blank.  I checked a few other tabs, all fine.  Back to Docker tab, still blank. Checked settings > Docker and it's telling me paths are invalid.  I have /mnt/cache/docker.img and /mnt/cache/appdata but neither seems valid at this point.

 

I checked the Dashboard, and it shows cache as inactive.  I checked under Main > Cache drives tab and is shows this...

 

Cache

KingDian_S280-480GB_2016082502509 - 480 GB (sdb)

*

23

0

0

xfs

Unmountable: Invalid partition

 

I'm not sure if this is an update issue, or a coincidental hardware issue, but before I risk doing further damage, I thought I should report it and see if there is a good way forward.

media-diagnostics-20170914-2326.zip

14 minutes ago, JustinChase said:

I just upgraded from rc7a and upon reboot, the array was not started.  I checked and saw disc 3 was empty.  I wasn't sure which disc should go there at first glance, so I just rebooted, and this time, it started the array. YAY!

 

However, I went to the docker tab and it was totally blank.  I checked a few other tabs, all fine.  Back to Docker tab, still blank. Checked settings > Docker and it's telling me paths are invalid.  I have /mnt/cache/docker.img and /mnt/cache/appdata but neither seems valid at this point.

 

I checked the Dashboard, and it shows cache as inactive.  I checked under Main > Cache drives tab and is shows this...

 

Cache

KingDian_S280-480GB_2016082502509 - 480 GB (sdb)

*

23

0

0

xfs

Unmountable: Invalid partition

 

I'm not sure if this is an update issue, or a coincidental hardware issue, but before I risk doing further damage, I thought I should report it and see if there is a good way forward.

media-diagnostics-20170914-2326.zip

 

Did you rule out the eariler posts in this thread regarding UD and having used that to format your the disk that you now have as your cache drive? Check that first and report back.

 

9 minutes ago, Lev said:

 

Did you rule out the eariler posts in this thread regarding UD and having used that to format your the disk that you now have as your cache drive? Check that first and report back.

 

no.  I don't remember using UD to format the drive, but it's certainly possible.

 

if so, what is the solution?  Is it possible to save the data, or am I hosed not that I've upgraded?  (I saw talk about the issue, but not any 'solution')

40 minutes ago, JustinChase said:

no.  I don't remember using UD to format the drive, but it's certainly possible.

 

if so, what is the solution?  Is it possible to save the data, or am I hosed not that I've upgraded?  (I saw talk about the issue, but not any 'solution')

 

You'd really best be served to re-read everything cause I can't promise you I'll recall it correctly.  

 

If you accept the risk that my suggestion is not as good as those before me in this thread...

 

• If your cache drive contains no data or data that is throw away, then pre-clear* it, then re-add as cache and format.
   
• If you do care about the data, I think the solution was to roll-back to the previous release, it should mount correctly again, back it up to another drive. Then perform the steps above above and restore from that backup.

 

*Pre-clear is not necessary required, it's just a means to wipe the drive to allow reformat that I find easy to use for that purpose. If you want quick and dirty then abort the pre-clear and proceed with format if your patience runs out before it finishes.

Okay, reverted to rc7a, moved everything off the cache drive via Midnight Commander, upgraded back to rc8, allowed unRAID to format the cache drive, copied everything back to the cache drive and rebooted.

 

Docker tab is blank and it's showing as stopped in the settings > Docker settings.  I disabled and saved, then re-enabled and it still shows as stopped, and the Docker tab is still blank.

 

I don't know what to do to get it started again.

 

Also, it takes about 30 seconds for the Plugins tab to load.  It started doing this on rc7a, and it's actually the main reason I upgraded to 8, hoping this was a bug that got fixed.  I don't know if this huge delay is now normal, but I preferred when it loaded rather instantly.

26 minutes ago, JustinChase said:

Docker tab is blank and it's showing as stopped in the settings > Docker settings.  I disabled and saved, then re-enabled and it still shows as stopped, and the Docker tab is still blank.

 

You probably have a corrupt docker image, delete an recreate.

 
Also, it takes about 30 seconds for the Plugins tab to load.  It started doing this on rc7a, and it's actually the main reason I upgraded to 8, hoping this was a bug that got fixed.  I don't know if this huge delay is now normal, but I preferred when it loaded rather instantly.

https://forums.lime-technology.com/topic/59417-plugin-list-takes-a-long-time-to-load/

12 minutes ago, johnnie.black said:

 

You probably have a corrupt docker image, delete an recreate.

 

I tried that shortly after posting, and it worked.  I had to reinstall all my dockers (only 5), but they kept their settings and download progress/history.  So, I guess I'm good now, other than the plugin tab taking too long to load.

Upgraded from rc7a. Now my WebGUI won't load anymore. I used the same IP address as before and now it won't load.

 

I can still access the server through SSH, it's just that the WebGUI is gone so I can't do anything now.

 

Can somebody tell me what to do? I can't get any diags because I don't have access.

SSH then type diagnostics.  Will collect and put them on your flash drive.

36 minutes ago, ideaman924 said:

Upgraded from rc7a. Now my WebGUI won't load anymore. I used the same IP address as before and now it won't load.

 

You likely have a port conflict with one of your dockers.

 

RC8 doesn't use the port settings in the go file anymore, but uses settings in the GUI instead. By default the GUI starts on port 80.

 

You can prevent docker from starting by editing the file "/config/docker.cfg" on your flash device and change the setting DOCKER_ENABLED to "No". Reboot your system afterwards.

 

2 hours ago, CHBMB said:

SSH then type diagnostics.  Will collect and put them on your flash drive.

 

OK, I've grabbed diagnostics. Attached below. Thanks for the command, didn't know that!

 

2 hours ago, bonienl said:

 

You likely have a port conflict with one of your dockers.

 

RC8 doesn't use the port settings in the go file anymore, but uses settings in the GUI instead. By default the GUI starts on port 80.

 

You can prevent docker from starting by editing the file "/config/docker.cfg" on your flash device and change the setting DOCKER_ENABLED to "No". Reboot your system afterwards.

 

 

Ah, just my luck. I'm still confused though. I do not have any Docker images on port 80. Will try with the trick you mentioned.

 

BTW, it was in /boot/config/docker.cfg, just so future users know.

 

EDIT: Nope it doesn't work. Times out for a short while (maybe 5-10 seconds after the server has booted) then it gives me a CONNECTION_REFUSED again.

derrickserver-diagnostics-20170916-2314.zip

 

 

EDIT2: I checked the logs, and this line is repeating over and over again:

 

Sep 16 04:01:23 derrickserver emhttpd: err: get_filesystem_status: Operation not supported (95): getxattr: /mnt/user/appdata-backup

Not really sure what is wrong but it repeats every millisecond and it also does it for /mnt/user/appdata too. Uhh...

Edited September 16, 20178 yr by Guest

@zin105 Did you accepted the untrusted certificate on the first execute? When yes and you don't close all session of the browser (chrome), then you don't get the untrusted certificate error anymore. With Firefox I thing you can add the untrusted certificate permanently to the certificate store. But in every case, add the URL line there should be a symbol for untrusted certificate.

Example:

Edited September 17, 20178 yr by fanningert

4 hours ago, zin105 said:

Accessing the box with https://HOSTNAME gives me an untrusted certificate, where as accessing it with http://HOSTNAME redirects me to https://HOSTNAME and is trusted. Is that intended?

 

@zin105Please excuse my quoting your post directly but I have an observation to make about the issue of Secure server access via free SSL Certificates from Let's Encrypt.  As near as I can determine the ability to use https: was first added in 6.4.0-rc3 with self-signed certificates and in 6.4.0-rc8q came the addition of 'trusted' certificates.  

 

There have been numerous problems and queries about these changes. Of course, all of these are now spread spread across four rc releases which makes for an additional confusion factor in finding the solutions to any problem that users encounter.  

 

Personally, I had/have an issue with provisioning my server with a trusted certificate because my router prohibits the operation because of security concerns.  (This is fixable in entering a series of commands on the router's command line window.)  I have elected not to modify my router at the present time because I am not convinced at this time that I really need a trusted certificate. (I suspect that 98% of all unRAID users are in the same situation.  I also wonder what will happen if the DNS server for this service goes down.)

 

What I think is needed is a separate thread to discuss SSL Certificate Settings, what additional safeguards each level (http:, https: with a self-signed certificate, and https: with a trusted certificate) provides, and then the problems and questions that users have.  Obviously, it would be ideal if the person (some LimeTech employee, perhaps) who starts this thread would have some real insight into the web security and its ins-and-outs.  That way, most of the questions, problems and solution will be in one thread.  It would be ideal if this could done before 6.4.0 is released.

 

9 hours ago, zin105 said:

That's beside the point. A certificate from Let's Encrypt should never show a red lock.

 

Not exactly. A cert from Let's Encrypt is valid only for the Fully Qualified Domain Name it was setup for, something like 6c4044c77838a17a5d5c58ddb2e63cbb7ad07b4d.unraid.net

 

If you try accessing it via a different hostname, such as:
  https://tower
then because "tower" does not match "6c4044c77838a17a5d5c58ddb2e63cbb7ad07b4d.unraid.net" you will get an untusted certificate. And because the cert is untrusted, your browser will not follow any redirect code it sends.

 

On the other hand, if you go to:
  http://tower
there is no encryption involved, and unRAID will be able to redirect you to the FQDN:
  https://6c4044c77838a17a5d5c58ddb2e63cbb7ad07b4d.unraid.net

1 hour ago, zin105 said:

 

Allright that makes sense. Original question still stands though.

 

Let's say a user goes to http://TOWER, get's redirected to https, creates a bookmark. Now every time they access the server they will have an untrusted certificate which defeats the point of the new implementation.

 

Sure they could look up why this happened and bookmark HTTP instead but that's bad design.

 

When a user goes to http://tower, they get redirected to the https unraid.net url. This url would not trigger the untrusted certificate and would be what gets bookmarked unless they manually change the bookmark to https://tower.

I experienced another freeze last night, however this time hard resetting the system was a bit of a gong show.   On the first hard reset it kernel panic'd on boot, then another reset and it KP'd during the array start on the third reboot i did safe mode and was able to get the array and containers up without plugins.  the parity check just finished and I have 6 errors apparently, i'll need to look into that after work.

 

These random freezes in the 6.4 series are very concerning, I haven't had such stability issues in a long time.  I'm wondering if it's the software or possibly RAM/USB key issues.  I'll try to keep an eye on dmesg for ECC errors.

Edited September 18, 20178 yr by Dephcon

11 minutes ago, Dephcon said:

I experienced another freeze last night, however this time hard resetting the system was a bit of a gong show.   On the first hard reset it kernel panic'd on boot, then another reset and it KP'd during the array start on the third reboot i did safe mode and was able to get the array and containers up without plugins.  the parity check just finished and I have 6 errors apparently, i'll need to look into that after work.

 

These random freezes in the 6.4 series are very concerning, I haven't had such stability issues in a long time.  I'm wondering if it's the software or possibly RAM/USB key issues.  I'll try to keep an eye on dmesg for ECC errors.

 

Have you turned on the Troubleshooting mode in the 'Fix Common Problems' plugin?  It might provide some clue...

Repeated server crash: 

 

 

https://forums.lime-technology.com/topic/60155-second-time-server-crash-on-latest-64/

 

WOW
Looking forward to this.

 

"Only thing" missing now is cluster support.

Would be way cool to cluster together unraid boxes

I can say I started using RC7 and 8 and never had a freeze or kernal panic once.

It's been up and running just fine since RC7 was released, with a few minor niggles like plugin load times, and 1 loose sata cable.

I get the plugin and os update check bug where you can no longer check for updates unless you restart the server. When I restarted it this time it also said I had an "unclear shutdown" even though I followed proper shutdown procedures.

On 11/09/2017 at 9:33 PM, Enver said:

Qemu / KVM issues?

 

Hi guys, running the 6.4 - RC8 and my main Windows 10 VM with GPU pass through is no longer working. Looking at the VM logs it suggests the GPU dump file I have used in previous versions is now failing to map? Please see below.

As a sanity check if I purposely change the file path to an incorrect path it does complain it can't find the dump file.

 

2017-09-11T11:22:50.079878Z qemu-system-x86_64: -device vfio-pci,host=05:00.0,id=hostdev0,bus=pci.4,addr=0x0,romfile=/mnt/cache/domains/Windows 10 Pro/evgagtx1080ftw.dump: Failed to mmap 0000:05:00.0 BAR 3. Performance may be slow

 

So I believe I found why my Primary GPU is not loading the driver properly; I found the following in the logs.

Sep 19 23:09:21 Tower kernel: vfio_ecap_init: 0000:05:00.0 hiding ecap 0x19@0x900
Sep 19 23:09:21 Tower kernel: vfio-pci 0000:05:00.0: BAR 3: can't reserve [mem 0xd0000000-0xd1ffffff 64bit pref]

Then I found this from Nvidia, see the 64-Bit BARs (Base Address Registers) section.

http://us.download.nvidia.com/XFree86/Linux-x86/173.14.12/README/chapter-09.html

 

What I can tell you is this particular VM with this XML and Syslinux.cfg(stubbed) has worked just fine with both 1080 GPUs being passed through to the Windows 10 VM. Ever since I upgraded to 6.4, I have not been able load the driver correctly for the primary GPU.

 

Can someone explain this; why has this worked in the past? Is anyone else having a problem passing through their primary GPU?

 

Thanks,

 

Enver