I updated to the latest version of UnRaid (v6.2.1) and Crashplan that runs ina docker container started misbehaving.
At first it kept cycling through “Syncronizing block information” and would not actually do a backup. It was stuck in a loop doing the sync operation.
I then changed the frequency of when the backup runs, by telling crashplan to run the backup between a specified time and days for all the backup sets.
That stopped the repeated sync operation, but now while doing backups, the Crashplan application (and engine) would crash and restart every minute.
I tried changing the -Xmx parameter in the run.conf file, but the restarting behaviour persisted.
I can monitor the docker container memory usage using the “cadvison” container, and I can see that as memory usage gets close to 2 GB, the system would restart that container. I bumped the -Xmx parameter all the way up to 4096MB, but it seems like it’s stuck at 2gb…
The next way I came across to increase the allowed memory usage is entering the CrashPlan console from within the app, ans issue the command “java mx 4096, restart”. (SEE LINK)
This successfully increased the allowed memory usage of CrashPlan to 4GB. Looking in the “cadvisor” app, I can see that crashplan now uses 3.5GB of ram, and everything is stable.
The webserver I was working on was running out of space only to find people had been uploading anywhere from 3Mb to 9MB jpeg images. Nobody needs that kind of image size and quality on the wordpress site.
Put together a script to for find and re-encode jpeg images over a certain size, to a lower quality setting.
I had this problem before where CrashPlan (the backup service) software will seem to be stuck “analyzing” data over and over without actually backing it up.
The last time that happened I had found no way around it except to re-start the entire backup. And since it takes months to transfer 1.8TB of data, it wasn’t a decision I tool lightly. I had searched all over the web and tried everything I could find, but to no avail.
This time I did some more digging again, and found a solution that worked.
CrashPlan for me runs as a docker container under UnRAID and UnRAID is a virtual machine in my ESXi homelab. I had 4GB allocated to UnRAID to do what it needed to do. The memory usage (at least as reported by unRaid) was about 60-80% so I dind’t htink I was running out of memory, but seeing as no other solution was working this time either I made 2 changes.
So the troubles seemed to be over with the ESXi server crashing when I was starting VMs that used PCI devices which had been passed through, but I was wrong.
The Win10 VM that had the GPU passed through is now once again either locking up the computer (ESXi locks up) or I get a PSOD (purple screen of death).
I read about others reporting similar issues and I updated the ESXi install to the latest build of version 6.0.0, but the problem persisted.
For reference, this is a listing of all the ESXi versions and updates and links to how to apply them (source).
One thing I just tried is disabling Interrupt Remapping (source) to see if it had any effect. It had no effect.
Removed and added the GPU device from the list of passed through devices.
Got my old Amiga 4000 out of storage at my parent’s place and decided to relive the glory days.
I still remember the great times I had with all the amigas I had owned. Starting with an Amoga 1000, then an Amiga 500, then an Amiga 2000 and finally an Amiga 4000. A few years after getting the A4000, I even remember dropping another $1000 on a Cyberstorm MKII 68060/50 Mhz accelerator card.
It must have been nearly 20 years since I powered up this machine.
There’s still a bit of testing to be done, but I may have gotten where I’ve been wanting to get for a while now in terms of virtualizing an remote desktop. That is the ability to use the GPU to render remote desktop sessions. By default this setting is turned off in Windows, and for the longest time I thought I had do crazy things like install VMWare Horison, or Xen Desktop (XDH) or even windows server to enable RemoteFX (which is the windows version of the GPU accelerated desktop delivery).
All along, it was just a few setting away in regular windows itself to enable RemoteFX.
Now running and playing on a remote desktop served by a virtual machine with a quadro 6000 gpu (well actually a modded gtx480 since it’s way cheaper then a real quadro 6000) in full HD.
Some testing is in order.
Bring on the heavy GPU use software!
What works so far:
– Full screen, full HD (1920×1080 video playback)