I have a Windows 10 guest that hangs at the BIOS screen. I have to force quit the VM, as trying to "Stop" or "Reset" causes the busy spinner to run forever. Parallels pins CPU at 100% for one core, and it never recovers even if allowed to run for an hour. I can recover the VM by rolling back to a much older snapshot. Rolling back to the last snapshot doesn't work, even if that was taken after a clean shutdown. But after a few weeks the problem will eventually happen again. It just happened to this guest for the fourth time in as many months. None of the other VMs on this machine have been similarly affected. Anything I can try the next time this happens? The hang is immediately after the BIOS shows the amount of memory allocated to Parallels. It never proceeds to the Windows 10 loading screen.
I've been able to reproduce this consistently now. I just have to create a couple snapshots of the VM and then it won't boot anymore. However, if I remove the snapshots it will boot again. Of note, this VM has two large (250GB) virtual disks attached, and each disk is configured for 2GB slices. This means Parallels may be opening around 250 files per active snapshot. Could Parallels be running out of file handles, or might it have a hard coded maximum number of disk slices that it can handle?
I can reproduce this 100% with a fresh VM now. I filed a support request (2608399) but don't know if that's the correct place to report bugs. In case anyone else can help usher it through, here's what I've filed. For now, I've consolidated my drives (not 2GB slices) as a workaround although it means less efficient backups for me. Bug Report: Reproducible hard lock with large number of drive fragments Reproduction: * Create a Windows guest virtual machine with a 250GB and a 256GB hard drive attached * Ensure the VM currently has no snapshots. Delete any that exist. * In the VM hardware configuration, for each hard drive, select "Properties" under advanced settings, check "Split the disk image into 2GB files" and apply. Conversion may take a few moments. * Boot the VM and confirm that it still works. Shut down again. * Open snapshots and take 3 snapshots. * Attempt to boot again Observed: * Guest VM locks up at the BIOS boot screen. Attempting to close or stop the VM will not work. Parallels continues to consume 100% CPU time. Expected: * Creating snapshots would not interfere with a normal boot Notes: * From here, you can kill the VM in Activity Monitor, delete the additional snapshots, and boot normally again. Alternatively, killing the VM then rolling back to the state of the first snapshot makes the VM bootable again. * From observing the VM with the fs_usage utility, it appears that Parallels locks up when it tries to open all the individual files that comprise all versions of the VM. There are about 253 files for the two virtual drives, plus an additional 253 for each of the snapshots. Parallels seizes up before opening the complete set. * Is Parallels increasing the number of file descriptors by enough to account for the additional drive slices? By default macOS allows 256 open files per process, but this value is adjustable. Search developer docs for 'maxfiles' for more information.
A support agent confirmed that they've escalated the ticket to engineering to be fixed in a future release. All's good.
