'mismatching index ID for us’ on almost every restart (again)

Same issue as https://github.com/syncthing/syncthing/issues/7051 / 'has mismatching index ID for us' on almost every restart

1.12.1 RC1

I can send a log, but I need to send a sanitized one, however every folder seems to be saying the same thing

I don’t see this on my devices and in the log you sent me there’s no repetition of a device+folder-ID, i.e. no indication that it does happen repetitively.

When do you see this happening?

Each time I restart. I will PM you an updated log which shows that each time I restart a new index is requested

Are you running an odd docker setup or something where the database isn’t persisted over restarts?

Pure windows setup via synctray. It’s on it’s own dedicated PC, no 3rd party AV. DB on an SSD.

apparently I had a panic - (panic-20201218-151345.reported), but that could be due to me restarting St is short succession

What is the panic?

Tow consecutive mismatches for the same device & folder combo:

129:[RTF25] 14:24:38 INFO: Device DWJVRMJ-... folder "A" (a) has mismatching index ID for us (0xA0175C8807ED7CBB != 0x42C01C3C03087F58)
[...]
566:[RTF25] 15:10:32 INFO: Device DWJVRMJ-... folder "A" (a) has mismatching index ID for us (0x42C01C3C03087F58 != 0x6F4E1206A392035F)

So the remote adjusts to the new id (sends it in the second exchange), but we locally already have a new one.

I don’t see anything out of the ordinary in between, that would explain it.

I think the way we get a new local index ID is racy, but that’s the case since forever and being a race, shouldn’t cause such consistent problems:
Both on sending and receiving a CC, we call fset.IndexID(protocol.LocalDevice) and in the beginning (no index id), both will create a new one. If both operations get a snapshot with en empty id, they will get two new, different index IDs.

@terry Would you be willing to test a binary from CI if I apply the fix for the race to v1.12.1-rc.1? On master: lib/db: Prevent IndexID creation race by imsodin · Pull Request #7211 · syncthing/syncthing · GitHub Not saying this is definitely the solution at all, but would like to know anyway :slight_smile:

Actually no need to test anything, the new log has the solution:

[RTF25] 14:24:34 INFO: Reinitializing delta index IDs

Nothing wrong with Syncthing, just -reset-deltas.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.