Capabilities
Checklist for Offsite Workshops
MSCF Resources
- Meet the MSCF Staff
- Acknowledging MSCF Use
- Help and User Tips
- Accounts
- Current Jobs Running
- Emergencies
- Glossary
- Journal Covers
- MSCF Brochure
- Ecce Brochure
- Training and Events
- FAQ
- Partners and Related Links
Additional EMSL Computing Resources
At least 2 months ahead of time:
- Figure out who onsite is going to be the contact person for Ecce and NWChem.
- Figure out who will be paying for and in charge of accomplishing what portions of the workshop (including travel expenses, copying, snacks for breaks, etc.) and get the arrangement in writing.
- Make sure license agreements have been signed for all software that is going to be used (including those that need to be purchased): NWChem, Ecce, ObjectStore (if the site is going to have a local database server).
- Get configuration of local machines and names of contact people with root access. Specific issues relating to Ecce that need to be discussed:
- oc_server is a root level daemon still, and is simply required on all servers. We can work better at providing convincing evidence of its security (or hopefully it will be done away with).
- Perl 5 is required for each platform, client and server.
- ObjectStore clients have some files that must be setuid root. This is an issue for some shared file systems (Ohio's policy was no setuid in nfs), affecting where it will be put.
- If they want their own ObjectStore server, they should ensure the machine has ample and/or expandable disk space, or it is hardly worth their/our time. Also, we need to make Ecce's requirements for authentication clear.
- Get accounts for the local machines (both compute servers and workshop client machines) for the instructors and any other support people.
- Set up workshop date, time and place.
- Set up a way to take registration (this may be someone who collects this information by e-mail or paper or it may be a web page).
- Send out announcement to those that are likely to be interested. This announcement should also be placed on the web somewhere.
At least 1 month ahead of time:
- Set up specific agenda for the workshop with times and instructors listed. This should be put onto the Web and sent to anyone who has or will register.
- Put together the workshop materials and send to the contact person offsite. (It is generally easier for the offsite contact to get copies made and bound. However, this is a detail that will need to be worked out.)
- Start getting all of the software installed and tested. This can take some time since every environment is different. DO NOT put this off until later.
- Specific issues related to Ecce:
- If it is a long-term installation, one person should read our documentation, read our documentation again, do the install while reading the documentation, and be responsible for user help. We can assist them, but we should not do all for them. Sometimes it is simply easier to do it for them.
- If they want globus, tell them we can't help them get it installed.
- Let them know individual workshop accounts need to be created on every workshop client and servers (unless users are assigned to a machine). These will need the correct environment set up, database accounts, etc.
- The ObjectStore client install requires a few key ownerships and one setuid (note setuid is an issue with shared file system policy!) else it will not work!: Must be owned by root: 1. $OS_ROOTDIR/lib 2. $OS_ROOTDIR/lib/oscmgr4 Must be owned by root and setuid: $OS_ROOTDIR/lib/oscminit
- oc_server uses the taint (-T) option, which guards against malicious use of a daemon script. Greg thinks he found the problem and fixed it. We shouldn't see it anymore, but, if oc_server just wont run, giving no errors at all, try turning off the -T taint flag in the first line.
- We no longer support running ecce from ksh or sh. It must be csh.
- For SGI machines, remember to tell them to ignore the "supportbad" error. May be gone in V1.5.
- Establish "reserved" queues for the class on the compute servers to be used. This may mean that Ecce and any NWChem launching scripts will need to be made aware of these queues. If possible, testing of these special queues should be done.
At least 2 weeks ahead of time:
- Close registration.
- Make sure that workshop materials have been copied and bound and are at the place of the workshop.
- Write any scripts that are necessary to do local launching of NWChem on compute servers. This is a back up to the launching in Ecce and should be tested by using several user accounts at the same time.
- Set up all of the users accounts so that they have identical environments (if possible), remembering to set up the .nwchemrc file to properly handle MD runs. The easiest way to do this is to set up a set of user files in a shared space that can be "sourced". Test each account using NWChem and Ecce (local launches and remote launches).
- Make sure that all instructors are aware of any potential problems with software and hardware.
At the time of the workshop:
- Have a great workshop!
- Take down any questions that were not answered during the class.
- Make sure the offsite contact is present and is introduced to the class.
- Remember to have all participants fill out an evaluation form.
After the workshop:
- Tabulate the results of the evaluation form and send to instructors, offsite contact, and any administrative people that should be informed.
- Find the answers to any questions that were not answered during the class and send them to class participants.