First, I declare that the data drives are USB, the indexes are on an SSD (c:\), it’s an i5 with 16Gb ram. Saying that, this version is the first where after a few hours, usually 4, I get the following under Synctrazor [image] I had the same issue last night and had to restart St, so I paused all …

[image] terry: […] is the hashing. Why though, did you reset the database? If it needs to rehash everything on upgrade something is wrong, that shouldn’t happen. I didn’t see any indication of that happening in the log you sent me previously either. Can you please take and share a cpu profil…

When I downgraded to 1.3.4 it wouldn’t talk to the upgraded indexes so I renamed the folder and let it reindex. Maybe i’m using the word hashing out of context. I mean scanning / indexing. After a day or so I had almost all the jobs reindexed and took it all back to 1.4. The upgrade took around 20 …

I’ve killed the current sync to enable debugging and here is the profile. Do you want one when it’s all locked up or at random times? syncthing-cpu-windows-amd64-v1.4.0-rc.5-095450.pprof (120.1 KB)

Another thing i’ve started to see, and don’t know if this is related or is a bug in itself, is the jobs say that are up to date, yet the local state differs to the global state. If I try a rescan on that job it comes straight back to being up to date. it’s not all the jobs, [image] and not relate…

Thanks for the cpu profile: Almost literally all accounted cpu time is spent generating folder summaries. Cpu usage should go down drastically if you close the web UI/Synctrayzor. I will file a PR shortly addressing this problem. As to why it seems worse than with 1.3.4 I have no idea. I can’t say …

This is now revolved, see Global state doesn't match local state on up to date folder

Not out of the woods. PC was dead in the water this morning, low CPU %, St on 3Gb ram, disk queue back in 60-70. No matter what I clicked on to open it, nothing happened, even outlook wasn’t syncing emails. Wasn’t able to get a cpu profile. After an hour trying I killed St and literally everything l…

To be honest, I think you should stop pointing fingers at syncthing and start pointing fingers at your drive. Sure, syncthing performs disk operations, but the fact that the rest of your PC locks up is usually a sign of a slow drive with high seek times. OS is usually smart enough to schedule IO re…

Where do you store the database - spinning disk or SSD? The timing makes me suspect the blocks GC as a culprit. On the other hand I don’t see why it would lock up your system. I checked the expected memory usage for your 1.5 million files with bloom package - github.com/willf/bloom - Go Packages an…

I am still sceptical. I think it’s just a dying drive that happens to be more exposed/visible because we do wider scans every now and then. There is nothing fundamentally wrong it what we do, if there was there would be more people talking about it. Again, run IO benchmarks, see if the numbers lin…

Syncthing Community Forum

Performance issues under 1.4.0 rc4

Support

show post in topic

Home
Categories
Guidelines
Terms of Service
Privacy Policy

Powered by Discourse, best viewed with JavaScript enabled