We have 2 instances of Nextcloud. The first instance has 2 active users (User1 syncing content across 3 PC and User2 syncing 1 PC with the server). The second instance has around 30 active users and roughly 50 in total (the only sync clients on this instance are for the same 2 users).
I’ll focus on the first one in this post, although the problem also happens in the second one.
We use NC to share our small office’s work documents. All files are stored into folder originally shared by User1. There are around 15,000 files & about 30GB of data. Some (most) of the data is various years old and has been untouch for at least 24 months (archived information from previous projects).
The instance was setup back when NC split from OC and has been upgraded various times throughout the years.
Starting around version 14, we started to see, from time to time, notices that there were conflicts with some of these archived files on both active users’ PCs. Checking the files in question, neither the original nor the the ‘conflict’ copy seem corrupted and the content for both is equal.
Since about 1 month ago, the problem has become much more frequent to the pont that I’ve started to seriously worry about the integrity of the data.
Before restarting the whole thing with a clean instance, I’ll like to see if anybody else here has run through similar problems and has managed to find a solution?
Our current setup (same for both instances):
Nextcloud 16.05.
Desktop sync clients: Version 2.6.0stable-Win64 (build 20190927).
Ubuntu 18.04, MariaDB 10.1, Apache 2.4 with PHP 7.3 (PHP FPM).
We use Redis for caching (although shouldn’t, I’ve been wondering if this could have to do).
Server is a dedicated box with 100 Mbps connection.
Users connected through 20 / 5 Mbps (Down / Up) connection.
Probably important to mention as well, the instance is replicated onto a backup server using MySQL replication + rsync of the nextcloud folder (excluding the config files).
I did try to pause the replication thinking it could be the issue but the problem has persisted.
I have seen this as well and you can find a few topics about this in the bug tracker https://github.com/nextcloud/desktop/issues. I didn’t find a good case how to analyse and debug this problem further, but it seems you did a bit of digging, therefore it would be great to let the developers know. Either join an existing report with the same problem, if it isn’t exactly the same or you have doubts, it’s better to open a new topic. Especially the observation that the frequency is increasing might help to identify the cause …
I also ran into similar issues when I started to test with clients that are doing VFS. Even though I am not using VFS at all, I think there are some side effects that bleed over to the standard sync code. It’s one of the reasons why I still use a very old owncloud 2.4.3 client. (Yes, I know, I probably should be using a Nextcloud client, but I’m currently afraid to move away from this really stable and working client.)
It’s a bit harsh, isn’t it?. There are three people who have seen this from time to time on a few files. There is no bug report, nobody can reproduce it (for the moment). As a developer, I’d start with confirmed reproducible bugs first.
@tflidd yes, it is harsh. And maybe it is unwarranted in this case. It’s just that this tendency to cram in new features, even though there’s so much potential to address core problems, is just mind-boggling.
Hi @SmallOne.
One user uses Windows 10 and syncs on 3 PCs.
One user uses Windows 10 and sync on 1 PC and one user uses OsX (not sure which version).
I’m wondering if the root of the issues is not the internet connection. The ISP we use is quite bad in general and it’s common for the connection quality to drop significantly with either extremely high latency or a ton of dropped packets. Unfortunately, the only 2 relevant ISPs in the country are both the same…
I haven’t had time on my hands to dig into this and try to see if there’s anything useful on the logs to spot the culprit. For the sake of everyone’s time, if you’d like, we can close this thread and I’ll reopen another one on the bug tracker link shared above if/when I have a clearer picture of the issue.
