Loading the 4GB saved state of a VM on a 13" MacBook Pro w/ Touchbar seems a bit slow: 20-70MB/s
Sequential reads/writes on the Flash Storage of this MBP go beyond 1000MB/s so why is the loading of a saved state so slow?
(Fusion 10.1.2, macOS 10.13.5)
This problem is gone since upgrading to MacOS 10.15.1 (Catalina) and VMware Fusion 11.5.0.
Fresh boot, FileVault on, 4GB VM loads in about 2 Seconds.
Upgraded to Mojave / Fusion 10.1.3 and it is still restoring at about 10-30MB/s on a brand new MacBookPro14,2 with APFS and FileVault.
Thing is, pretty much the only thing running on this Device is VMware Fusion. No other non-Apple Software that could interfere.. so is this normal?
Does this have something to do with FileVault?
Yes, without FileVault, it's much faster: fresh-reboot restores at about 200-300MB/s.
But since this is a "mobile" device, running around without FV is not really an option..
There were some reports that APFS does this. The workaround is to put the VM in a sparse bundle container that's HPFS formatted.
I'm having the same problem with Fusion 11.0.2 on Mojave. I'm using an iMac with 32GB RAM, with around 14GB given to a Windows 10 VM (which has been running, saving and restoring fine until very recently). I don't use FileVault on this iMac. And none of my VMs have snapshots.
When I restored the Win10 VM today it took 13 minutes.
I created a sparsebundle disk image (formatted Mac OS Extended (journaled)) and moved the VM to that. There was an improvement, but a restore still took around 6 minutes.
I also have an Ubuntu VM that showed the same problem. Shutting it down and then rebooting seems to have solved the problem for now. But doing the same with the Windows 10 VM (and I have tried several times) had no effect.
I've been having the same issue ever since upgrading to Mojave too - I too believe this is due to APFS. After several days of back and forth with VMWare support, they could not help. I felt that perhaps my Mac was just in need of a refresh since I have been upgrading since Yosemite. So I did a complete format and restore from scratch. This seemed to have fixed it and I was very hopeful, but this week one of the VM's starting taking at least 15 minutes to restore, without warning. It is the most used one and gets suspended every night and restored every morning. It is my belief that after X number of suspend/restore actions that this problem arises and does not ever go away afterwards. It is to the point that it is faster to just shut down (power down) the VM every night and start it up (power up) every morning. IMO, this is ridiculous and should not be taking this long to rectify.
I suspect this is related to the new approach to time machine that leverages APFS snapshots. If you disable time machine, reboot, and try resuming again, does that fix the problem?
I do not use Time Machine....
Another clue in case anyone at VM Ware is reading this.....
If I cancel a restore process then immediately launch it again, it will rapidly restore up to the point where I cancelled it then slow down back to a snail's pace again..... very frustrating - have been waiting for a restore to finish now for nearly 10 minutes and it's only about 30% done.
Just spent 40 minutes waiting for my VM to resume because I forgot to shut it down before restarting my Mac, so it automatically paused the VM - guess it's time to try out Parallels. VMWare just isn't providing me the experience I so loved when I was on a Windows PC, or that I had pre-APFS. I don't think they know what is wrong or simply don't believe there is a problem, or just can't fix the darned thing. I've had 2 updates since originally bringing this to their attention and the problem still persists. I LOVE VM Ware, but it is practically un-useable at this point.
Fusion 11 seems to have fixed the APFS related issues a while ago.
One thing that could be causing the problem is if the VM is consuming more RAM than is available in the system when you restore. That causes the OS to page out while Fusion is reloading the memory image.
I have approx. 4X the VM's memory available at any given time. I've already gone around and around with support checking all of these things but they wanted to remote in to my machine and I simply cannot allow that due to the sensitive client data I have. Everything worked flawlessly until Mojave (that's the first version where fusion drives got converted to APFS). I even did a complete reformat and reinstall of the OS and all apps. The VM's worked good for the first 15-20 resumes then slowed back again and have never gone back to "normal". This is problematic on numerous VM's not just one.
Ok, that's interesting. Couple of things it might be:
- Make sure 'autoprotect' and 'linked virtual machines' aren't enabled.
- If you use time machine, make sure that the VM folder is excluded from it, and that the time machine volume is attached to the computer. If you don't use time machine make sure it's actually 'off'
The first could be a fusion issue. the latter is a design flaw in how time machine uses snapshots under APFS. I'd be curious to know if you manually remove all local snapshots, if the resume time improves. the terminal command is tmutil listlocalsnapshots / (assuming you're on the internal SSD) to find them, and then tmutil deletelocalsnapshots <date> for each one to delete them.
If you use Carbon Copy Cloner, disable CCC snapshots on the impacted volume too.
- Make sure 'autoprotect' and 'linked virtual machines' aren't enabled.
Auto-protect is off on 1 but on on another - both have the same issue - only 1 snapshot for the machine with no auto-protect - but I have tried deleting them before to no avail
No liknked VM's anywhere
- If you don't use time machine make sure it's actually 'off'
I don't use time machine - it's turned off - never runs.
- If you use Carbon Copy Cloner, disable CCC snapshots on the impacted volume too.
I use CCC and always have snapshots disabled - they make restoring cumbersome
Ugh - the only other idea is that if you use a Fusion drive, it may be shifting the VM off to the metal disk, but short of that, I'm out of ideas.
Just want to make a note of this. On 7-10-2019 I completely uninstalled and re-installed VM Ware. So far resumes are performing as they should. Time will tell if another complete re-install will be necessary after X number of resumes. They worked fine back when I completely re-installed MacOS and all apps until X number of restores, at which point they starting stalling again, so I'm assuming that will happen this time too. I will update here accordingly.
Just a note - 7-15-2019 - restores have slowed down somewhat. After a fresh uninstall and reinstall of VMWare on 7-10-2019 they were running around 5 seconds. Today they are taking 30-50 seconds.
7-25-2019 - Today's restore took over 4 minutes.
Now an interesting observation from yesterday.... yesterday morning's restore took under 2 minutes. After using for 3-4 hours I paused it. About 30 minutes later I remembered something I had forgotten to do, so I resumed it and it took less than 5 seconds. I left it on the rest of the day and paused it before I left for the evening. This morning I got to work and the resume took over 4 minutes.
Just as before when I did a complete machine reinstall, it slowly gets longer and longer until I will eventually need to uninstall and reinstall VMware. I foresee this being a once per month event at the rate it's going right now.
This is just bizzare. If you were using a Fusion disk, I could understand it, but otherwise it just doesn't make sense.
Long shot but have you tried timing other large sequential disk reads over time? i.e. copy large file to the drive? Wonder if there's something going on at the disk/OS level.