[JinZhou5042]

No description provided.

[btovar]

Hi @JinZhou5042, why is by apparent ip needed? I don't think we needed that before with glados.

[JinZhou5042]

It's because glados advertises two addresses: 10.32.93.156 and 169.254.1.2 and workers try them one by one.

Workers sometimes hang trying 169.254.1.2 that the condor nodes can't reach. by_apprent_id makes them use only 10.32.93.156

[btovar]

What do you mean hang? Don't they eventually try the next ip?

[JinZhou5042]

Yes they do, but each try on that ip address takes about 2 min before the connection times out. And it seems that each worker may try multiple invalid addresses in sequence so that delay can add up.

[btovar]

Sounds good, thanks! Were you able to run all of it after these fixes?

[JinZhou5042]

@btovar Yes, after these fixes along with cooperative-computing-lab/vine-cms-analysis-stack#1 I was able to run everything. I'll start a new run in a fresh environment to make sure everything works correctly after fixing these issues.