Q. My submitted job is still in the queue - why is it not running?
A. There could be several reason why your job is not running:
If you have access to the members.q queue, it could be that you and other users in your lab are currently using all your slots, which in case your jobs are being queued in the communal long.q queue instead.
The queue where your job is sitting may be full. If so, your job will eventually run.
You might have asked for compute resources that cannot be met,
e.g. more memory or more cores than available on any compute
node, e.g. -l mem=4048gb or -pe smp 256. If so, your job
will never run. Either lower the job’s resource needs using
qdel, or, alternatively, remove the job (qdel) and submit
(qsub) a new one with adjusted resources.
If there is a downtime scheduled, you might have asked for a run-time that goes into the downtime period. If so, your job cannot be started until after the downtime. If you don’t specify the run-time when you submit your job, the default is 14 days.
qstat -j <job_id> will provide details on why a particular job
is not running. qstat -u '*' will show all jobs and their
priority scores in the queue.
A: [For QB3 migrants] If you are coming from the legacy QB3 cluster (defunct since 2020), please make sure that your job script does not specify any of the below QB3-specific SGE resources. A job submitted with one or more of these will sit in the queue forever.
-l arch=linux-x64: this architecture does not exist on Wynton HPC. This specification can safely be dropped on
Wynton HPC. (You could specify, -l arch=lx-amd64
but that is not needed as all compute nodes now have the same
architecture.)
-l database=<size>: this storage resource does not exist on Wynton HPC.
-l netapp=<size>: this storage resource does not exist on Wynton HPC. This specification can safely be dropped on
Wynton HPC.
-l scrapp=<size>: this storage resource does not exist on Wynton HPC.
-l scrapp2=<size>: this storage resource does not exist on Wynton HPC.
Q. I tried to delete some jobs, and now they’re stuck in the state “dr”. How can I get rid of them?
A. The most likely cause of this is that node (or nodes) running your jobs crashed. Since the node(s) can’t report back to SGE and confirm the job deletion, the state of the jobs doesn’t change. To force the issue:
qdel -f $JOB_ID [-t $SGE_TASK_ID]
Q. What is the difference between the idgpu, iogpu, and
atgpu parts used for GPU compute node names?
A. They denotes CPU architecture: io is for “Intel
Octo-core” (i.e. Intel CPUs with 8 cores per CPU), id is for
“Intel Dodeca-core” (12 cores per CPU, but now encompasses all
Intel nodes with more 12+ nodes), and at is for “AMD
Triginti-core” (32 cores).
Q. I just started to get SSL-related errors when using qsub and
qstat that I have never seen before;
error: commlib error: ssl connect error (SSL handshake error)
ssl error (the used certificate is expired)
unable to contact qmaster using port 6444 on host "q"
A. Your Wynton HPC account has expired. If so, you should already have received an email from us with instructions on how to request the renewal. If you have responded to that email, then it’s a mistake on our end (sorry) - please drop us another email.
Q. x2go authenticates, but then immediately disconnects without launching.
A. The first thing to check is your BeeGFS home directory quota. x2go creates a number of temporary files related to your sessions. If it cannot create those files, it will authenticate and then disconnect. Short version: beegfs-ctl --getquota --storagepoolid=11 --uid "$USER". For more information on quotas and file systems, see the page on File Sizes and Disk Quotas.
Q. How do I terminate a saved x2go session if it seems “stuck” or “unresponsive”?
A. Log in to the destination server you were connecting to (not the proxy/jump host) via ssh, type x2golistsessions.
If something comes up, run x2goterminate-session $SESSION, replacing $SESSION by the second field in the output of x2golistsessions.
Then try connecting again with your x2go client.
Q. I am getting timeout errors when trying to connect via X2Go from a macOS computer, the X2Go status hangs on “connecting”; In the X2Go logs you will see:
Info: Forwarding X11 connections to display '/private/tmp/com.apple.launchd.C24DSqSnIF/org.xquartz:0'.
Info: Forwarding auxiliary X11 connections to display '/private/tmp/com.apple.launchd.C24DSqSnIF/org.xquartz:0'.
Session: Session started at 'Tue Mar 2 13:00:37 2021'.
Connection timeout, abortingSession: Terminating session at 'Tue Mar 2 13:01:05 2021'.
Info: Waiting the cleanup timeout to complete.
Session: Session terminated at 'Tue Mar 2 13:01:07 2021'.
A. This appears to be a communication problem between X2Go and XQuartz. The only way we’ve found to resolve this issue is to Completely remove XQuartz from the macOS computer and then re-install XQuartz. Please follow recommendations for completely removing the XQuartz application and all related files. (Search for any files or folders with the program’s name or developer’s name in the ~/Library/Preferences/, ~/Library/Application Support/ and ~/Library/Caches/ folders.) After re-installing XQuartz, X2Go should work again. If not, please contact the Wynton team.
Q. I tried to change my shell using the unix command chsh and I
got an error telling me, “chsh: user ‘alice’ does not exist”.
A. First, let me assure you, your account does exist! You are
logged in, after all. However, Wynton account
attributes are managed via a remote directory system which is not
manipulable via local tools like chsh. If you would like to change
your shell, Please get in touch with the Wynton
team, let us know your preferred shell, and we will change it for
you. Note: The Wynton team supports csh/tcsh and
sh/bash login shells. Any other shell than these may result in
reduced functionality or errors which may be beyond the scope of our
support.
Q. I cannot SSH into the development nodes - I get ‘IT IS POSSIBLE THAT SOMEONE IS DOING SOMETHING NASTY!’ and ‘Host key verification failed.’ What is going on?
A. This most likely happens because we have re-built the problematic development node resulting in its internal security keys having changed since you last access that machine. If the problem error looks like:
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
@ WARNING: REMOTE HOST IDENTIFICATION HAS CHANGED! @
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
IT IS POSSIBLE THAT SOMEONE IS DOING SOMETHING NASTY!
Someone could be eavesdropping on you right now (man-in-the-middle attack)!
It is also possible that a host key has just been changed.
The fingerprint for the ECDSA key sent by the remote host is
SHA256:FaARLbkE3sHP2a33Zgqa/sNXTqqWzZAeu6T43wST4ok.
Please contact your system administrator.
Add correct host key in /wynton/home/boblab/alice/.ssh/known_hosts to get rid of this message.
Offending ECDSA key in /wynton/home/boblab/alice/.ssh/known_hosts:18
ECDSA host key for dev1 has changed and you have requested strict checking.
Host key verification failed.
then the solution is to remove that offending key from your personal
~/.ssh/known_hosts file on Wynton. If you get
this error when you try to access, say, dev1, then
use:
$ ssh-keygen -R dev1
to remove all SSH keys associated with that machine. Alternatively, you can manually remove the problematic key by looking at:
Offending ECDSA key in /wynton/home/boblab/alice/.ssh/known_hosts:18
to identify that we want to remove the key on line 18. To remove that line, use:
$ sed -i '18d' ~/.ssh/known_hosts
Then retry.
Q. Why do I get “incorrect password attempts” when using sudo
despite entering my password correctly?
A. The sudo command is only available to system administrators.
It is a command used to run a specific software as root, that is, with
administrator privileges. First of all, for security reasons, users
do not have the rights to use sudo. Second, sudo is often used
to install a software tool centrally on the current machine for
all users. If you think about it, it would wreak havoc if any user
would be able to install or update software that other users use.
By the way, if you ever get prompted for your Wynton HPC password, please stop and think! Except for when you access the cluster, or one of the development nodes, no software should ever need to know you password. So, if you get asked for your Wynton HPC password, do not enter it.
Now, if you do call sudo, the system administrators will be
notified automatically, and you will most likely get a follow-up email
from them. Please respond to such a message, if you get one. That
said, if you end up using sudo, press Ctrl-C when you are
prompted for your password. This will prevent the command from being
completed, e.g.
$ sudo make install
We trust you have received the usual lecture from the local System
Administrator. It usually boils down to these three things:
#1) Respect the privacy of others.
#2) Think before you type.
#3) With great power comes great responsibility.
[sudo] password for alice: <Ctrl-C>
$
Q. Is data on Wynton backed up?
A. Data on Wynton is not backed up! Users and labs are responsible to back up their own data outside of the Wynton HPC environment.
Q. Is it possible to have a common folder where our lab group members can share files and software?
A. Labs who purchase additional storage will get a
/wynton/group/your_group/ folder. Files written in that folder will
not count toward users disk quota.
Q. When I run Jupyter, I now get an error saying
/opt/rh/rh-python38/root/usr/bin/python3: bad interpreter: No such
file or directory. How can I fix it.
A. This is because the Python software, here Jupyter, was
installed using the rh-python38 SCL, which was only available on
CentOS 7. To fix this, you need to reinstall the broken software,
which in this example means reinstall Jupyter.
Q. I might have corrupted my Bash startup files. How do I reset them?
A. To get a fresh ~/.bashrc file, make a backup of your old one
and copy the default one by:
$ cp ~/.bashrc ~/.bashrc.20231115
$ cp /etc/skel/.bashrc ~/
A. To get a fresh ~/.bash_profile file, make a backup of your
old one and copy the default one by:
$ cp ~/.bash_profile ~/.bash_profile.20231115
$ cp /etc/skel/.bash_profile ~/
Q. Can I use Microsoft VS Code with Wynton?
A. Yes, but, importantly, do not use Microsoft Visual Studio
Code (VS Code) on the login nodes. Because of this, you need to
configure VS Code on your local machine to connect directly to a
development nodes instead. To achieve this, start by configuring your
local SSH environment to connect directly to a development node via a
“jump host” directive as described in ‘Connect directly to a
development node’. When you know that works, consult VS Code’s
documentation regarding using an SSH “Jump Host” or “Ssh Proxy”. In
short, configure VS Code by clicking on the SSH extension, “new
remote”, and then enter ssh dev1. This should allow VS Code to
connect directly to dev1.
Q. Why don’t you like screenshots of output and error messages?
A. We, and others, strongly prefer that you cut-and-paste textual output from your SSH terminal in your email, Slack, … messages. There are several reason for this. First, and most importantly, screenshots are not accessible, which means they are useless to a person using a screen reader. For others, text in screenshots might be very hard to read and require zooming in on the image. Second, it is not possible to cut-and-paste from a screenshot, which adds unnecessary friction to anyone trying to reproduce your problem and help you. Third, contrary to plain text, screenshots are not searchable in email clients and on Slack.
Q. Our lab would contribute to Wynton HPC in order to increase our priority. How can we do this?
A. We welcome donations of any size. In return, your lab will receive a number of slots in member.q equivalent to the number of cores in a current Standard Node that your contribution would purchase. As of May 2020, that cost is 170 USD per slot.
Q. Our lab has some old nodes we’d like to contribute to Wynton HPC in return for priority. Will you take them?
A. Please get in touch with the Wynton team.
Q. Does Wynton Provide Server Hosting, Data Storage, or Application Hosting Services?
A. Wynton HPC provides High Performance Computing resources to the UCSF Research community. We are not able to provide server hosting, application hosting, data storage, or consulting services outside of our core mission in Research High Performance Computing. For these, and related services, we suggest you contact UCSF IT regarding their offerings, e.g. UCSF IT Virtual Server Hosting, UCSF IT Physical Server Hosting, and UCSF IT Cloud Services.