Edge Workspaces sync BUGS

[Arti-G]

Hello
First of all, I think that Workspaces in Edge is a very useful feature! Edge also has vertical tabs, which makes it the best tool for researchers using hundreds of tabs. I think other browsers can learn from this. With Workspaces, I can move windows I don’t currently use aside and open them again when needed without loading to RAM. Other browsers lack this feature, so I moved to Edge from Chrome. Edge is now my main tool for ongoing online research.

However, after some months of use, I see that there are bugs that make me frustrated. With Workspaces, Edge is slow, it freezes, fails to synchronize, and reverts changes.

1. It does not update tabs and moves new tabs to the end.

Let’s say that I have opened tabs: New Tab, facebook.com, A, B, C, D, E

Later, I open links in new tabs from New Tab, Facebook, and A and leave them open. The tabs are: New Tab, msn.com/F, facebook.com, G, A, A2, B, C, D, E

After some time, Workspaces synchronize, there is a short or longer Edge freeze. Then I see tabs are reorganized and F, G, H moved and changed addresses. The tabs are now: New Tab, facebook.com, A, B, C, D, E, New Tab, l.facebook.com/l.php?u=G, A2

It looks like Workspaces never update opened tabs when there was redirection through New Tab or l.facebook.com/l.php?u= or any other redirection on other websites. It only stores the first address it sees. When I refresh these tabs to see the description and go to the proper addresses, it still reverts it back! After I organize the order properly, it sometimes stays or is reverted back! This is frustrating when you need order and proper descriptions!

2. It is terribly slow, especially after opening another Workspace.

I also encountered long freezes of Edge when opening another Workspace window or at random, probably when it synchronizes (a red dot appears next to the workspace name). Sometimes I have to wait 5 minutes, sometimes 15 minutes, sometimes I’m losing patience! I have plenty of PC power (12-core Ryzen 3900X, 32GB RAM, Samsung SSD 970 EVO Plus 1TB) and when I check Process Hacker, I see that Edge isn’t using 100% of CPU, but still is frozen. I noticed that when it’s frozen, there are around 200 disk writes, fast changing, 9MB/s total:
C:\Users\me\AppData\Local\Microsoft\Edge\User Data\Default\Sync Data\LevelDB\256443.ldb ... ( edge_workspace_disk.xls )

When LevelDB writes stop, Edge is unfrozen.

I’ve given msedge.exe processes high priority, but it doesn’t solve anything.

I noticed this on previous versions starting from 122. I use Stable and Beta (to have two separate profiles) on PC and laptop. I waited hoping it will be solved, but nothing changed. It's possible that developers only tested Workspaces with a limited number of tabs and not with power users like myself, who manage 10 workspaces with 200-600 tabs in each. The consistent unreliability suggests there may be unresolved bugs present in each version.

I examined the LOG.old file in the Sync Data>LevelDB folders of each browser and noticed that when a freeze occurs, there is a specific behavior logged:
2024/06/12-13:14:19.485 1a570 Generated table #365187@0: 40283 keys, 2109396 bytes

2024/06/12-13:14:19.485 1a570 Compacted 8@0 + 1@1 files => 2109396 bytes

2024/06/12-13:14:19.488 1a570 compacted to: files[ 4 6 7 0 0 0 0 ]

2024/06/12-13:14:19.489 1a570 Delete type=2 #365171

2024/06/12-13:14:19.489 1a570 Delete type=2 #365172

2024/06/12-13:14:19.489 1a570 Delete type=2 #365174

2024/06/12-13:14:19.489 1a570 Delete type=2 #365176

2024/06/12-13:14:19.489 1a570 Delete type=2 #365178

2024/06/12-13:14:19.489 1a570 Delete type=2 #365180

2024/06/12-13:14:19.489 1a570 Delete type=2 #365182

2024/06/12-13:14:19.489 1a570 Delete type=2 #365184

2024/06/12-13:14:19.489 1a570 Delete type=2 #365186

2024/06/12-13:14:19.491 179a0 Level-0 table #365197: started

2024/06/12-13:14:19.492 1a0a0 Current memtable full; waiting...

2024/06/12-13:14:19.500 179a0 Level-0 table #365197: 467718 bytes OK

2024/06/12-13:14:19.503 179a0 Delete type=0 #365194

2024/06/12-13:14:19.504 1a570 Level-0 table #365199: started

2024/06/12-13:14:19.505 1a0a0 Current memtable full; waiting...

2024/06/12-13:14:19.511 1a570 Level-0 table #365199: 467718 bytes OK

2024/06/12-13:14:19.514 1a570 Delete type=0 #365196

2024/06/12-13:14:19.515 179a0 Level-0 table #365201: started

2024/06/12-13:14:19.516 1a0a0 Current memtable full; waiting...

2024/06/12-13:14:19.524 179a0 Level-0 table #365201: 467718 bytes OK

2024/06/12-13:14:19.530 179a0 Delete type=0 #365198

2024/06/12-13:14:19.530 1a570 Level-0 table #365203: started

2024/06/12-13:14:19.531 1a0a0 Current memtable full; waiting...

2024/06/12-13:14:19.539 1a570 Level-0 table #365203: 467718 bytes OK

2024/06/12-13:14:19.542 1a570 Delete type=0 #365200

2024/06/12-13:14:19.542 179a0 Compacting 8@0 + 1@1 files

2024/06/12-13:14:19.556 179a0 Level-0 table #365206: started

2024/06/12-13:14:19.564 179a0 Level-0 table #365206: 467718 bytes OK

2024/06/12-13:14:19.568 179a0 Delete type=0 #365202

2024/06/12-13:14:19.572 179a0 Level-0 table #365208: started

2024/06/12-13:14:19.581 179a0 Level-0 table #365208: 467718 bytes OK

2024/06/12-13:14:19.584 179a0 Delete type=0 #365205

2024/06/12-13:14:19.588 179a0 Level-0 table #365210: started

2024/06/12-13:14:19.596 179a0 Level-0 table #365210: 467718 bytes OK

2024/06/12-13:14:19.600 179a0 Delete type=0 #365207

2024/06/12-13:14:19.604 179a0 Level-0 table #365212: started

2024/06/12-13:14:19.611 179a0 Level-0 table #365212: 467718 bytes OK

2024/06/12-13:14:19.615 179a0 Delete type=0 #365209

2024/06/12-13:14:19.620 1a0a0 Too many L0 files; waiting...

3. Workspaces sometimes stop syncing randomly and without notification.
Occasionally, I notice a mismatch in the workspace between the PC and the laptop I use. If I close the workspace without checking, I will loose the tabs opened since the last sync! There's no notification, I must check if everything is synced or manually compare the opened tabs in the same workspace and then open them manually. After arranging them in the original order, I return to bug #1, which causes a mess when Edge moves tabs around.

4. Issues with Accessing Workspaces.
This issue has occurred to me several times and is common among many users: "Error: Unable to load workspaces." Edge Workspaces error V114 - Microsoft Community Hub I tried many workarounds, sometimes they worked, yet they are not ideal and a definitive solution remains elusive.

Now please tell me that Workspaces aren't still in beta phase. Workspaces are generally excellent tool, but in their current state of development, they are just a joke not worth the frustration and risk of data loss.

[#2]

Somehow Edge is not frozen now, not doing hardcore writes to SSD, so I checked logs and compared old (frozen) to current log (working ok).
Not frozen:

2024/06/14-01:23:19.445 8b00 Compacting 4@0 + 2@1 files
2024/06/14-01:23:19.471 8b00 Generated table #594894@0: 24741 keys, 2108148 bytes
2024/06/14-01:23:19.491 8b00 Generated table #594895@0: 39630 keys, 804534 bytes
2024/06/14-01:23:19.491 8b00 Compacted 4@0 + 2@1 files => 2912682 bytes
2024/06/14-01:23:19.492 8b00 compacted to: files[ 0 4 7 0 0 0 0 ]

Frozen 1:

2024/06/12-13:14:19.485 1a570 Generated table #365187@0: 40283 keys, 2109396 bytes
2024/06/12-13:14:19.485 1a570 Compacted 8@0 + 1@1 files => 2109396 bytes
2024/06/12-13:14:19.488 1a570 compacted to: files[ 4 6 7 0 0 0 0 ]

 Copilot says:

Comparing the two logs, there are a few differences that might have contributed to the improved performance:

1. Number of Files Compacted: In the second log, 4 Level-0 files and 1 Level-1 file were compacted. In the first log, 4 Level-0 files and 2 Level-1 files were compacted. This means more data was being processed in the first log, which could potentially lead to better organization and faster access times.

2. Number of Keys and Size of Data: In the second log, one table was generated with 40,283 keys and a size of 2,109,396 bytes. In the first log, two tables were generated with a total of 64,371 keys and a combined size of 2,912,682 bytes. The distribution of keys across multiple tables and the larger total size of data could potentially improve the performance of read operations.

3. Number of Files After Compaction: In the second log, there were 4 files at Level-0, 6 files at Level-1, and 7 files at Level-2 after compaction. In the first log, there were no files at Level-0, 4 files at Level-1, and 7 files at Level-2. Having fewer files at Level-0 and more files at Level-1 could potentially improve read performance, as LevelDB tries to minimize the number of Level-0 files due to their overlapping key ranges.

Frozen 2:

2024/06/13-11:56:15.168 7118 Compacting 4@0 + 1@1 files
2024/06/13-11:56:15.425 7118 Generated table #542179@0: 64278 keys, 1575665 bytes
2024/06/13-11:56:15.425 7118 Compacted 4@0 + 1@1 files => 1575665 bytes
2024/06/13-11:56:15.558 7118 compacted to: files[ 0 3 7 0 0 0 0 ]

Copilot says:

Comparing the two logs, there are a few differences that might have contributed to the improved performance:

1. Number of Files Compacted: In the second log, 4 Level-0 files and 1 Level-1 file were compacted. In the first log, 4 Level-0 files and 2 Level-1 files were compacted. This means more data was being processed in the first log, which could potentially lead to better organization and faster access times.

2. Number of Keys: In the second log, one table was generated with 64,278 keys. In the first log, two tables were generated with a total of 64,371 keys. The distribution of keys across multiple tables can potentially improve the performance of read operations.

3. Size of Data: The total size of the data compacted in the first log is larger than in the second log. This could indicate more efficient packing of data.

4. Number of Files After Compaction: In the second log, there were 3 files at Level-1 after compaction, while in the first log, there were 4 files at Level-1. Having more files could potentially improve read performance, as it’s easier to locate data spread across multiple files.

[#3]

It seems that no dev cares. I hope this isn't the case, but it appears that after reading my post, it was simpler to ban one of my MS profiles in Workspaces, or it's just #4 bug. On my main PC, I have access to my list of workspaces and the freezes have ceased, allowing me to use it normally. However, on my two other laptops and the laptops of two friends, I'm unable to retrieve the list of workspaces. Attempts to resolve the issue, such as reinstalling Edge, Edge Beta, and deleting user data, have been unsuccessful, indicating the problem is not on my end. I fear that if I close Edge, I will lose all my workspaces. If anyone is willing to address this, I am available to assist as a bug tester.

[#4]

@Arti-G 

After month of troubleshooting and 0 help from Edge developers here are my findings:

Regarding #1 bug nothing changed and I can't do anything. Workspaces have to understand that there are 200 then JAVASCRIPT redirect from Facebook and Instagram (maybe other sites too).

Regarding #4 bug with problems to access your workspaces you have to close your account and create new one, sorry. Problem is on Microsoft servers. Sometimes you might access your workspaces from other browser or computer but it is random and you will loose access eventually. If you have this in your edge://workspaces-internals :

Spaces Event: EdgeFluidCSPStorageHelper failed at OnGetContainerListComplete

{ "Date": "0", "ErrorCode": -102, "PrimaryError": 17, "SecondaryError": 37 }

Spaces Event: WorkspaceManager: Failed to initialize Fluid Error code:46 tag:-41

Spaces Event: Fluid Manager Process Failed

"Failed to initialize Fluid"

Just run! Microsoft won't help you and won't send you a backup. If you want to feel safe create new account and use Tab Session Manager that will save sessions everytime you close browser or window.

Regarding #3 bug I have no idea why Workspaces can stop sync without a notice. Sometimes it can stop sync, but there will be red dot. 

And most important #4 bug:
It appers that Workspaces is poorly written and highly abuses Chromium sync (made for syncing bookmarks, settings, history, not for workspaces) with LevelDB in Sync Data folder. Opening any Workspace start up to 3000 I/O writes delta and 2000 I/O reads delta, writes to disk are also high making hundreds of gigabytes a day. It contributes to SSD wear! It doesn't freeze a system and there are resources left to run even another browser, but such I/Os make browser unresponsive. I'm sure it's because workspaces because I opened same set of tabs without workspaces and there was no problem and I also saw that when I disconnect form the Internet or block IPs connecting with Edge - it's unfrozen - workspaces can't sync and LevelDB writes go low. Workspaces make LevelDB tables with many keys and long values of favicons and updates fast. LevelDB can't keep up so it freezes the browser.
Can anyone except developers do something about it? Only idea I think is to move Sync Data folder to Ramdisk, I have to test it. If it fails then workspaces are just useless, full of bugs crap with no support. 

[#5]

I'm continuing to work on optimizing Workspaces on my side since I have not received a response from the developers and there are no changes in the new versions of Canary. Workspaces are useful but constantly freeze, making it unusable. If I cannot make it function smoothly, I will stop using it.

I installed ImDisk and created RAMdisk for 'Sync Data' folder. It moved I/O from SSD to RAM, it makes writes much faster and prevents from excessive wear Edge does to SSD. Unfortunately, it does not stop Workspaces from experiencing hours-long freezes.

I think there is a limit of CPU power allocated for one process. The main Edge process does not utilize full CPU power when necessary. On my two PCs, it utilizes either 12.5% or 4.1% of CPU power. This correlates with the number of threads on these PCs—12.5% for 8 threads and 4.1% for 24 threads. It appears that the main process is using only one thread (12.5*8=100%, 4.16*24=100%), although the process manager indicates it is using more than one. Is there a limit to CPU usage for a single process? Can I allocate more CPU to the main process to enhance its speed?
The screenshot from Process Hacker shows the CPU usage for the main process is flat, as if it was restricted in some manner. The CPU has 24 threads, suggesting it may have been limited to 1/24th of the CPU's capacity. However, it works across multiple threads.