LeMill blog

Last week our storage for media pieces maxed out at ~32000 pieces, preventing uploading of new pieces. During this week we have reorganized storage to use directory schema that allows practically unlimited amount of pieces and all of the pieces are now copied to new storage and the storage is activated. Uploading pieces should be working.

This change could also lead to some kind of speed gains, as we don’t have to deal with one huge directory anymore.

Teachers in Georgia have especially been very productive with great photos and images and we hope to see the flow of great resources continuing after this unfortunate pause.

Our hosting service found the cause of late performance problems and LeMill.net should now be much more comfortable to use.

There still is an issue of us having too many great media pieces, and the directory where they are stored cannot hold them anymore. This requires us to reorganize the storage, which will take few hours some night when there isn’t much activity.

Hello, and apologies for late slowdowns in LeMill. We have been working really hard to fix them, but unfortunately my efforts have often been misguided: I have found some slow operation that seems to be the cause and found out how to speed this operation on our test server, and when we finally have the fixed version running in actual LeMill site, we have found that the service is still as slow as it was — the problem was elsewhere.

Now we have extremely professional people from LeMill’s hosting service looking at the issue, and as he has much more better view on how the data moves in our system, he can probably soon point where the problems are.

In mean time, I’ve found two possible causes or minor causes of slowdowns and fixed them.

Making the problem worse is that the slowdown is at its worst when restarting the service, so activating our fixes at daytime is often not worth the uncertain benefit.

I hope I can get some of the problems cleaned during tonight 15.12-16.12, 23:00 EET onwards.

We had to pull LeMill into read-only mode, as we fix some catalog issues. Service should resume in a few hours, our goal is 14:00 EST (GMT+2) today.