[v1.6.1] Local and global state swapped between two nodes


I’ve got strange mismatch between two receiveOnly nodes receiving the same stuff from one SO node. Is this interesting to debug together or should I take the hammer-hit and reset the database to help myself out?

If interesting, I would take the time to do a debug run again.

The devices and folders all show “up2date” (green status) about each other.

device A : SO It shows the correct amount of files in the folder = 2.814 / 540 / ~ 166 GiB


device B : RO (switching to SR does show the same picture) image

device C : RO (switching to SR does show the same picture) image

Kind regards, Catfriend1

Could you copy the dbs and then take the following steps, noting if anything changes about the global/local states:

  1. One after the other run the RO nodes with -reset-deltas.

  2. Run with STRECHECKDBEVERY=1s to check/fix db and recalculate meta.


Hi @imsodin,

I took the following steps:

  1. Shutdown devA, devB, devC
  2. Copied databases
  3. Started devC with -reset-deltas

All folders show like this: image

  1. Started devB with -reset-deltas

The affected folder showed 2.814 local and 2.814 global files as state for some seconds and then showed the error state printed above (2.838 g, 2.814 l)

devC shows 2.814 g, 2.838 l (still error state like in the inital post)

  1. Started devA with -reset-deltas

It still shows 2.814 g, 2.814 l (like in initial post)

  1. Stopped devA.

  2. Put “export STRECHECKDBEVERY=1s” in front of Syncthing start on devB, devC.

  3. Restarted devC. No change in g/l accountings. Log showed “Stored folder metadata for “…” is 442376h9m21.825384227s old; recalculating” lines.

  4. Restarted devB. No change in g/l accountings. Log showed “Stored folder metadata for “…” is 442376h11m5.898206352s old; recalculating” lines.

  5. All in all, the problem persists.

Do you need the database copies?

Foot notes:

  • On all devices, I saw the line “Reinitializing delta index IDs” coming up in the log after starting with the command line switch.

Hmm I don’t know how to go from here but the outcome of syncing seems still fine. Just the report looks strange on the UI.

Ups, this somehow escaped my attention. The state persisting is definitely weird. Yes, I’d be interested in having a look at those dbs (though no promises…).

1 Like