Examples not working?

Frequently Asked Questions (FAQs) #

Jobs #

Waiting in job queue #

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:

  1. 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.

  2. The queue where your job is sitting may be full. If so, your job will eventually run.

  3. 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.

  4. 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.

Cannot delete jobs from queue #

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]

What do the different hostname prefixes stand for? #

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).

Errors #

Cannot submit jobs #

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.

X2Go does not connect #

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.

chsh: user ‘alice’ does not exist #

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.

Scary error when trying to log in to a development node #

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 dev2 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, dev2, then use:

$ ssh-keygen -R dev2

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.

Don’t use sudo #

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>

$ 

Files and folders #

Are my files backed up? #

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.

Share folder with group members? #

Q. Is it possible to have a common folder where our lab group members can share files and software?

A1. If you belong to a specific group, we can set up a /wynton/home/your_group/shared/ folder that group members (part of the same Unix group) have write access to. Any such files will count toward the disk quota of the user who owns the files. The typical use case is then that one or more members maintain subdirectories therein. If you need this, please drop us an email. Note, if the groups command reports lsd for you, then you do not belong to a specific group and can unfortunately not get a group-specific folder.

A2. 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.

Miscellaneous #

Reset my shell startup file #

Q. I might have corrupted by Bash startup file. How do I reset it?

A. To get a fresh ~/.bashrc file, make a backup of your old one and copy the default one by:

$ cp ~/.bashrc ~/.bashrc.20220912
$ cp /etc/skel/bashrc ~/.bashrc

Contributing to Wynton #

Purchase compute slots #

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.

Contribute hardware #

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.

What Wynton offers #

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.