High CPU load sustained overnight - mac - build 6271

Came back from being gone overnight with my odrive @ cpu usage >100%. Quit and relaunched odrive, and it is now hovering around 40%, and has been for ~4 hours.

While my odrive icon in the tray only shows ‘ready to sync new changes,’ it is switching from black to the pink ‘activity’ mode. I did install anaconda into a google drive folder, but it looks like all of those binaries and packages have synced.

Thoughts?

EDIT: Interesting - I synced my anaconda directory to another computer. The ‘pkgs’ directory synced, but then many of the packages became .cloudf files and began to sync again.

I stopped automatic sync, on the computer where I installed odrive initially, and it shows ~500 files waiting, all of which look like they originate in the pgks anaconda directory. I don’t have any terminal sessions running.

Thoughts?

Hi @alex.hakso,
Apologies for the late reply. I missed this post.

Are you still seeing this issue? I’m thinking that perhaps the anaconda folder has some symlinks that could be causing some issues? If this is still happening, can you send a diagnostic from the odrive tray menu?

It was definitely the anaconda folder, and the pkgs directory in particular. I moved the installation outside of my google drive, and the problems went away. I suspect that, if you have the interest, you could replicate this immediately by doing a standard anaconda install into a google drive folder.

Thanks @alex.hakso!

The QA team is running through this scenario and we’ll see what it yields.

Great. I would be curious to know the results, if you remember

It is a large, sprawling structure:
Files = 82262
Directories = 7173
Symlinks = 342

It will take a while to make it through, but I am guessing that the land mines may be in the symlinks. We try to follow them, so if any are wonky it can cause problems.

Damn, that is a lot to crawl through for a standard installation.

Obviously the sprawl and number of files has a high initial cost to get synced, but wouldn’t necessarily cause ongoing problems. Do the ongoing issues definitively indicate bad architecture in the anaconda installation, or are there some sound architectures that just don’t work for cloud syncing?

We haven’t seen the issue yet with the anaconda structure, but its got a ways to go.

The scale of a structure has a large impact on initial sync, as you said. It also has sustained, periodic impact because odrive will continue to closely monitor the entire structure for changes, both local and remote, which can create quite a bit of overhead. We are working on ways to optimize this, but unsyncing less-used/unused portions of data also goes a long way in mitigating the effect.

When there are symlinks in play, they are usually my primary suspect for problematic structures. Generally the result of a self-referencing symlink, or a broken reference.