I have slow and huge folder 5M files ,it was in initial scan when update to 1.4.1 appears, restarted and now folder in weird state ,
2020-04-07 10:21:04 Failed initial scan of sendonly folder “AA” (cfcyx-rwk6a)
state was “unknown”, then I paused resumed, it state scanning, but nothing happens. that state on 2 computers.
I cant simply delete database ,as a few more folders involved, the are seems OK.
How to deal with it ?
You see other errors on the folder in the GUI explaining the problem. The log entry is just a summary.
Please don’t do that, there is no reason and it won’t help.
I may panicking 
I did restart one node , seems nothing at all in the log file that correlated to the error … scanning resumed.
i guess scanning should pass the point where it was before to starting reflect increasing numbers of files? in my situation approximately 3-4 days as it already shows 3.6M files.
too bad there is no real time indicators of the process … “scanning” one node , “scanning 0%” other , that 2 hours pass…
The GUI does have real-time progress for the scan, once it’s counted the files that it needs to scan.
not sure about real time, when it in routine re-scan daily, it just state “scanning” blue then “up to date” green , and there is no any indicator of process. initial scan shows some percentage. re-scan - nothing …
like existing folder currently in scanning process: Files:210,534 Folders:43 ~80.1 GiB , state just “scanning” … no any information of the process …
if you have lots of (small) files it is possible that Syncthing only shows “scanning” without ETA or progress indication for quite some time.
Rescan shows no progress if nothing has changed, the work % estimated is based on the stuff that has changed.
would be nice to have an additional line of information …
like in my state: it indicated Files:3,263,894 Folders: 7,851 ~627 GiB And i guess it half way done, so I may expect a few days “scanning” without any information , and if it not stuck maaabe some percentage finally appears ?
I saw memory spikes up to 3.8GB on a previous release. Lucky i have 12GB, If I not, and the software fail to allocate more memory, would it be exception, failure, code misbehavior, application restart database corruption? This indicator disappear now… Minimalistic design?
Sorry, I’ve read your post but I am not sure what your question is, please be more specific.
OK … Back to the problem: Problem cause: Initial scan was interrupted by self-upgrade. Shared folder , approximately 5-6M Files, 1.2TB. Slow source.
Problem state: “SCANNING” - nothing reading from a datastorage ever…
Folder was left on “SCANNING” mode, now it “UNKNOWN” mode. update do 1.4.2 Folder from “UNKNOWN” mode back to “SCANNING” mode. Task manager for shows zero activity in reading\writing; same amount bytes for 20 min now. while folder in “SCANNING” mode …
GUI log blank , completely … this is write only node for this folder:
data from text log file:
INFO: Ready to synchronize “RS” (cycnh-rfk2w) (receiveonly)
INFO: folder “RS” has mismatching index ID for us (0x3DB1AA2DA7B92F35 != 0x370AB06CE8A97054) INFO: folder “RS” has a new index ID (0x14D8B032A6E9AC98)
so far related to folder
What does the task manager show you actually? In the scanning phase it is indeed not reading anything, it is just looking up metadata for your entire file tree. In linux that would show up as a high %iowait value on syncthings process. No clue what’s the equivalent in task manager on windows, maybe you can look at “overall load” of the hard disk, which should be close to 100%.
Anyway it doesn’t sound like anything is wrong: If your source is incredibly slow and large, you’ll just have to wait (or change your setup).
Again: “SCANNING” state - zero byte read or write by application. Any access to the file system (db, logs, anything …) reflecting as data read or write in bytes.
restarted over web interface, crashed
Current state: one folder scanning finished: 220K Files , 100GB. Task monitor indicated: 9.8GB read 1.3GB write after scanning this folder.
second folder “SCANNING” zero activity in read or writing … no any changes …
now what ?
…I think it corrupted … inconsistent database. I need some instruction to clear mess, starting from beginning for this folder or clear coruption …
No. Accessing file system metadata results in 0 bytes read and 100% load.
Logs please.
For that too, logs please.
UPD: enable some debug, related to database:
repeating this line every min or so… ; nothing more going on:
[RLNX7] 2020/04/08 12:34:51.766000 set.go:184: DEBUG: cycnh-rfk2w WithNeedTruncated(7777777-777777N-7777777-777777N-7777777-777777N-7777777-77777Q4) [RLNX7] 2020/04/08 12:35:23.657600 set.go:184: DEBUG: cycnh-rfk2w WithNeedTruncated(7777777-777777N-7777777-777777N-7777777-777777N-7777777-77777Q4) [RLNX7] 2020/04/08 12:35:58.181600 set.go:184: DEBUG: cycnh-rfk2w WithNeedTruncated(7777777-777777N-7777777-777777N-7777777-777777N-7777777-77777Q4)
Sure, I might be wrong about windows filesystem, not ashamed to admit that, and I am happy to be pointed out as wrong. However I can tell you that you better don’t expect to get any help with an attitude like this where you ignore most of the advice given and react with a link to some general documentation article.
no offense pal, I’m really appreciated you willingness to help. I got SUN Solaris Admin. cert. in 97; and MS engineer a few years later , when grass was green, and Oracle not gobbled everyone, and so I clearly know a differ how it looks in diff. kind OS. That Why I’m stated, “EVERYTHING”
I suggest you start syncthing with STTRACE=scanner env var set, and capture the logs for some amount of time until it’s been hanging on scanning for 5mins or so (rather than just a few lines), and upload that somewhere for us to look at. Trying to selectively pick a few lines from the logs usually does not give us what we want.
got it, in process…
UPD: Question: should I keep some debug enabled, log with “scanner” not so talkative, just 28 lines? if debug need to en. , what logging facilities should turn ON ?
You should enable it via env var as I’ve explained, and capture logs via stdout, not logs from the web ui, as those show last 100 lines or so.