Continued from thread 8857 (re moving BOINC from SSD to HDD)

Message boards : Questions and problems : Continued from thread 8857 (re moving BOINC from SSD to HDD)
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57056 - Posted: 25 Oct 2014, 5:41:08 UTC
Last modified: 25 Oct 2014, 5:51:20 UTC

Here is the final post of thread 8857. Unfortunately (but understandably) I am not able to respond directly to it.

I'm back on the case. My SSD had a sudden failure. I've replaced it, and then a couple of weeks later the number of concurrent tasks started to drop, so that means that I should just give BOINC loads of space on the HDD.

I've managed to get the HDD to mount at startup, which was preventing the move previously.

Now to get the BOINC data folder and its permissions, and those other configuration items, sorted out. Not today though.
ID: 57056 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 57075 - Posted: 26 Oct 2014, 20:39:29 UTC - in response to Message 57056.  

So you have the HDD mounted at /media/el_gallo_azul/Storage. I recommend moving the mount point somewhere else.

When you mount other partitions, USB sticks, CDROMs or anything, GNOME uses a directory under /media/el_gallo_azul/ as mount point for the device. By default /media/el_gallo_azul has permissions set so that only you can access stuff under it. I believe you use distro's BOINC which means boinc user will need access past /media/el_gallo_azul. While you could loosen the permissions I wouldn't trust GNOME to not reset the permissions at some later time.

So mount the HDD elsewhere. /mnt/storage is good. Your attempt to use it failed for two reasons: Disks doesn't create the mount point for you, and Disks adds some mount options that the rest of the system doesn't understand. Since Disks doesn't work you need to do everything yourself.

Create the mount point and adjust permissions:

sudo mkdir /mnt/storage
sudo chmod o+rx /mnt
sudo chmod o+rwx /mnt/storage


This allows everyone to write to /mnt/storage. Drop the 'w' if that is not what you want.

Next edit the mount point in /etc/fstab. Reboot and check /mnt/storage with file manager to make sure the HDD is mounted.

Before moving BOINC's data directory stop the client.

sudo service boinc stop


Use System Monitor or ps to make sure the client and all the science apps have stopped.

Then move the data directory. The steps below also reset ownership and permissions. They are probably unnecessary but I included them just to be sure.

sudo mv /var/lib/boinc /mnt/storage/boinc
sudo chown -R boinc:boinc /mnt/storage/boinc
sudo chmod u=rwx,g=rwx,o= /mnt/storage/boinc


NVIDIA drivers at least at some point stored something (kernel cache maybe) in user's home directory. There maybe some other programs that expect users to have home directories. The boinc user has it's home directory set to /var/lib/boinc. To keep everyone happy place a symbolic link at the old location pointing to the new location.

sudo ln -s /mnt/storage/boinc /var/lib/boinc


With the symbolic link in place you don't need to adjust BOINC's configuration.

And finally, cross your fingers and restart boinc service.

sudo service boinc start



(I didn't actually test these instructions. You need to fix typos yourself.)


As for the Dropbox, you'll need to either tell it the new location of it's folder or add a symbolic link from /media/el_gallo_azul/Storage pointing to /mnt/storage.
ID: 57075 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57077 - Posted: 26 Oct 2014, 21:40:37 UTC - in response to Message 57075.  

Thanks Juha.

I was going to make a post to the thread yesterday (that I had moved /var/lib/boinc-client to the HDD, created the symbolic link, checked that permissions had been kept for the boinc-client directory, and updated /etc/default/boinc-client, but that BOINC still didn't work) but boinc.berkeley.edu seemed to be down most of the day.

I suspected that the next thing to tackle would be the permissions, and then luckily I saw your post before I changed anything. What you say about /media/el_gallo_azul/ makes sense.

I'll remount the HDD to /mnt/ (etc) later today. It looks quite promising.
ID: 57077 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57094 - Posted: 28 Oct 2014, 5:15:23 UTC - in response to Message 57075.  
Last modified: 28 Oct 2014, 5:54:09 UTC

I'm getting closer.

sudo mkdir /mnt/InternalHDD

sudo chmod o+rx /mnt

sudo chmod o+rwx /mnt/InternalHDD/

- In /etc/fstab file, change

"UUID=4f721fc0-8072-453f-b48f-ca686bd89549 /media/el_gallo_azul/Storage ext4 defaults 0 2"

to

"UUID=4f721fc0-8072-453f-b48f-ca686bd89549 /mnt/InternalHDD ext4 defaults 0 2"

-Reboot
- Check that /media/el_gallo_azul/ and /mnt/ look OK

- Uh oh
sudo /etc/init.d/boinc-client stop
gives "* BOINC data directory '/media/el_gallo_azul/Storage/boinc-client' does not exist."

sudo chown -R boinc:boinc /mnt/InternalHDD/boinc-client

sudo chmod u=rwx,g=rwx,o= /mnt/InternalHDD/boinc-client/

sudo rm /var/lib/boinc-client
(to delete previous link from /var/lib/boinc-client to /media/el_gallo_azul/Storage/boinc-client) then

sudo ln -s /mnt/InternalHDD/boinc-client/ /var/lib/
(to create new symbolic link)
- Edit /etc/default/boinc-client file with BOINC_DIR="/mnt/InternalHDD/boinc-client"

- Reboot
- Try to start BOINC Manager. The BOINC Manager icon flashes 6 times (ie. about 12 seconds), but that's all that happens.

Happily,
sudo /etc/init.d/boinc-client start
(or stop) or
sudo service boinc-client start
(or stop) work OK (although I won't be able to add any projects until the Manager is running).

Dropbox was easy to fix, but VirtualBox will be a bit of fun.
ID: 57094 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57095 - Posted: 28 Oct 2014, 6:39:11 UTC - in response to Message 57094.  

I had a quick brainwave, so I started boincmgr from a terminal. It starts, but in the terminal I get the message

"execvp(/home/el_gallo_azul/boinc, --redirectio, --launched_by_manager) failed with error 2!"

Also, the BOINC Manager application shortly (within a minute or less) gives the message

"The BOINC client has exited unexpectedly 3 times within the last 2 minutes. Would you like to restart it again?"
ID: 57095 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57096 - Posted: 28 Oct 2014, 7:24:15 UTC - in response to Message 57095.  

OK I left my computer for an hour or so, and restarted it a little while ago. I wanted to copy and search for the error text when starting boincmgr (I didn't try starting BOINC Manager from the Launcher icon). This time, the application started OK, and I have been able to add a project. I don't know if it's working yet, because of Status "Communication deferred 00:59:59", so I'll add some more projects and see if it tries to start calculating.
ID: 57096 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57097 - Posted: 28 Oct 2014, 7:28:02 UTC - in response to Message 57096.  

I've added a couple more projects, and just checked what was happening in the terminal. This has appeared so far:

16:47:55: Error: Memory VFS already contains file 'webexternallink.xpm'!
16:47:55: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
16:47:55: Error: Memory VFS already contains file 'atiicon.xpm'!
16:54:38: Error: Memory VFS already contains file 'webexternallink.xpm'!
16:54:38: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
16:54:38: Error: Memory VFS already contains file 'atiicon.xpm'!
16:55:29: Error: Memory VFS already contains file 'webexternallink.xpm'!
16:55:29: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
16:55:29: Error: Memory VFS already contains file 'atiicon.xpm'!
ID: 57097 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57098 - Posted: 28 Oct 2014, 7:46:24 UTC - in response to Message 57097.  

Despite those messages, I've added some other projects, tasks have downloaded, and the tasks have now started "Running".

Happily, I have 32 tasks running again, whereas before this move I was down to 23 concurrent tasks.

I still can't be 100% sure that it's working until these work units are completed and uploaded, of course, but it's looking pretty good.

The /mnt/InternalHDD/boinc-client folder has all the correct subfolders etc:

http://imgur.com/ID5TM7v
ID: 57098 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57099 - Posted: 28 Oct 2014, 7:47:17 UTC - in response to Message 57098.  

From the terminal (so far):

16:58:57: Error: Memory VFS already contains file 'webexternallink.xpm'!
16:58:57: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
16:58:57: Error: Memory VFS already contains file 'atiicon.xpm'!
17:01:00: Error: Memory VFS already contains file 'webexternallink.xpm'!
17:01:00: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
17:01:00: Error: Memory VFS already contains file 'atiicon.xpm'!
17:01:40: Error: Memory VFS already contains file 'webexternallink.xpm'!
17:01:40: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
17:01:40: Error: Memory VFS already contains file 'atiicon.xpm'!
17:02:32: Error: Memory VFS already contains file 'webexternallink.xpm'!
17:02:32: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
17:02:32: Error: Memory VFS already contains file 'atiicon.xpm'!
17:03:16: Error: Memory VFS already contains file 'webexternallink.xpm'!
17:03:16: Error: Memory VFS already contains file 'nvidiaicon.xpm'!
17:03:16: Error: Memory VFS already contains file 'atiicon.xpm'!
ID: 57099 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57100 - Posted: 28 Oct 2014, 8:01:10 UTC - in response to Message 57099.  

PS It's still running, and I have Tools > Computing preferences... > disk and memory usage > Use at most 256 Gigabytes disk space and left everything else as the BOINC Manager default.
ID: 57100 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 57182 - Posted: 28 Oct 2014, 21:49:03 UTC - in response to Message 57097.  

I've added a couple more projects, and just checked what was happening in the terminal. This has appeared so far:

16:47:55: Error: Memory VFS already contains file 'webexternallink.xpm'!
16:47:55: Error: Memory VFS already contains file 'nvidiaicon.xpm'!


These are normal for 7.2 Managers, nothing to worry about.
ID: 57182 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57187 - Posted: 28 Oct 2014, 22:24:53 UTC - in response to Message 57182.  

Gotcha. Thanks.

I don't know what happened, because between Message 57095 (not working) and Message 57096 (seemingly working) I didn't change anything, but BOINC Manager (and BOINC) still appear to be working. Maybe the computer just needed that second restart.

Since Message 57096, it seems like there have been a higher proportion of "Computation error", and some tasks with a Remaining (estimate) of over 7 hours have been "Ready to report" within about 10 seconds or so of starting to run, but I'll leave BOINC Manager running for a couple of days and see what happens on boincstats (for example).

Thanks very much for your help - particularly Message 57075.
ID: 57187 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 57189 - Posted: 28 Oct 2014, 23:00:02 UTC - in response to Message 57187.  

Glad you got it working.

Better check for problems right away.

Over at CPDN, you have a few tasks marked with Client detached. Messing with the client may have caused those, hard to tell.

At POGS, two task with "finish file present too long". Depending on your point of view, that's either a bug or a feature in the client possibly combined with a buggy app.

And finally, at GPUGRID you have a few tasks with "SWAN : FATAL : Cuda driver error 3". That looks like a bug in their app.
ID: 57189 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57191 - Posted: 28 Oct 2014, 23:54:44 UTC - in response to Message 57189.  

Yes I just noticed another CPDN task "Computation error". I have had about 4 or 5 instances of "Computation error" with CPDN before. These tasks seem to be a week old, which is when the work units were happening on my SSD.

I noticed the "Computation error" last night mostly for POGS tasks, which is unusual. They have always been OK before.

I haven't installed Cuda so I have never had GUPGRID working 100%. I have only completed GPUGRID tasks that require only CPU.
ID: 57191 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57194 - Posted: 29 Oct 2014, 2:29:28 UTC - in response to Message 57191.  
Last modified: 29 Oct 2014, 2:57:42 UTC

Update: A couple of tasks have appeared from yesterday (for CPDN errors). Hopefully, they're the same type of error that I used to see before I moved from SSD to HDD for work units.
ID: 57194 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 57206 - Posted: 29 Oct 2014, 19:50:26 UTC - in response to Message 57194.  

"terminate called after throwing an instance of 'std::bad_alloc'"

Memory allocation failed. Either your host ran completely out of free memory or the app tried to allocate some silly amount of memory.
ID: 57206 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57207 - Posted: 29 Oct 2014, 20:45:15 UTC - in response to Message 57206.  
Last modified: 29 Oct 2014, 21:00:13 UTC

I'm just looking at the problem now.

I've just posted to a CPDN Message Board about it as well.

I just looked at an example of a work unit for which I had had a "Computation error", and it looks like all of its tasks have had errors on a variety of computers with different operating systems. I am computer 1326407 for that work unit.

It would be a shame for me to run out of free memory. I got several times what I (thought that I) needed, but maybe I should have gone straight to 128GB (the vendor's maximum) right from the start. Or maybe even 512GB.

Hopefully these errors are happening because of some quirk with the current batch of CPDN work units.
ID: 57207 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57214 - Posted: 30 Oct 2014, 1:20:15 UTC - in response to Message 57207.  

I had to restart my computer after a kernel update earlier today, so I took the opportunity to see if BOINC Manager started from the Ubuntu launcher icon now. It doesn't work - again, like in Message 57094, the BOINC Manager icon flashes 6 times (ie. about 12 seconds), but that's all that happens, so I have to start it from a terminal using
boincmgr
.
ID: 57214 · Report as offensive
Profile el_gallo_azul

Send message
Joined: 19 Jan 14
Posts: 79
Message 57219 - Posted: 30 Oct 2014, 9:35:54 UTC

I have a couple of completed CPDN tasks showing now, so it all looks good. That's that one done. Hooray!
ID: 57219 · Report as offensive
Juha
Volunteer developer
Volunteer tester
Help desk expert

Send message
Joined: 20 Nov 12
Posts: 801
Finland
Message 57233 - Posted: 30 Oct 2014, 21:27:30 UTC - in response to Message 57214.  

It's been a while since I lasted looked at how the Manager is set up to start in Debian. I think by default it starts in /var/lib/boinc-client. Since you now should have the symlink there, question is have you modified the icon/.desktop-file to start the Manager somewhere else?

Your account also needs to be a member of boinc group, I think.
ID: 57233 · Report as offensive
1 · 2 · Next

Message boards : Questions and problems : Continued from thread 8857 (re moving BOINC from SSD to HDD)

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.