It's been a busy week since I started the blog series on the new LSF Web Service interface ( part 1, part2 & part 3) with a lot of downloads, some great feedback, and quite a few questions.

I've summarised the questions in a mini-faq below, and will add to as required.

LSF Web Service - FAQ

1) Is there a webinar on July 31st?

Maybe!   The event registration functionality built into this site only shares your username with me - not your contact details, so I can't send you information on how to download the package.    So I'm using webex as a registration mechanism to be able to contact you.   I'm planning on running the preview until the end of the July.  If any significant changes are needed to the package, I'm reserving the 31st for an AMA.

2) Can I call Support ?

No, not for this technical preview at present. Either just post in the discussion group, or email us directly. Once we formally release it, it will be supported via usual IBM channels.

3) Will this be available on IBM Cloud?

Yes, we will be adding it to the LSF offerings on IBM Cloud.

4) Will you support ARM/Graviton?

We've already added an aarch64 (client and server) package to the preview.

5) Will you support MacOS (on Arm) ?

It's in the backlog, we plan to add it.

6) Is there a cost for this add-on?

It's not an add-on -- it will be part of LSF Standard / LSF Suites - so no additional cost.

6a) Really?

Yes!

7) Can we redistribute the client side with our application?

It wasn't something we have considered at this point. It would require an update to the LSF license agreement itself, but it's a plausible idea - leave it with us.

8) Can we integrate this with our corporate single sign on?

SSO support will be added later in the preview process - please let us know which SSO solution you are using.

9) Is it easy to migrate existing LSF scripts?

Yes, you can just add "lsf" before each existing LSF command in the script. The web service returns output in exactly the same format as the CLI so any scripts that parse the output should require no modification.

If you don't want to modify the scripts, you will shortly be able to create symbolic links to the "lsf" client, for which every LSF commands you want locally. e.g.

$ ln -s lsf bsub

$ bsub sleep 100

10) How do you handle output that is larger than the web service maximum request length?

• Like all web services, there is a maximum request length (typically 4MB) to protect against denial of service attacks that request huge amounts of data. In our case, rather than a data size, we've set it at 500 lines of output ensuring a complete record is returned, rather than something chopped off in the middle due to a data size limit. This limit is configurable by the the administrator for the web service.
• If you are using the provided "lsf" client, and the output is more than 500 lines, the full output is directed to a zipped file on the server, which is then automatically retrieved by the client, and displayed to the user.
• If you are using the API directly and the command returns more then 500 lines, the the first 500 will be returned to the user, and the full output will be directed to a zipped file on the server, which can then be downloaded.
  
  

------------------------------
Bill McMillan Principal Product Manager, IBM Spectrum Computing
------------------------------