Iāve been pushing my luck for a while running beta macOS software (started with Big Sur) and itās finally catching up to me I guess. Funnily enough some months ago I had stopped using beta software and thatās when my external NVMe booting setup stopped working. After I then temporarily switched to the internal HDD and later an SSD I foolishly jumped on the Beta bandwagon again. What happened is I signed up for Beta to see if that makes a difference with the NVMe boot troubles but I never got a Beta that time and I forgot about it. Much later though when I started using an NVMe again (details in that previously linked post) I was offered and update and chose it by habit without realizing it was a Beta one (as I still had that Beta profile installed).
That stopped Karabiner Elements from working and I worked around that. The Karabiner issue got fixed in Beta 2 but that oneās been particularly worse for me. I run it on my iMac and it crashes every day (or at least every other day). I canāt find any crash logs but I get a message about the kernel and invalid addresses ā sort of like this guy on the Twitters. I couldnāt find any other reports of this so maybe itās just me and that guy. I donāt use Chrome though, only Firefox. I donāt think I have any crazy software installed either ā just āstandardā stuff like Visual Studio Code, iTerm2, etc. It is very frustrating though when it keeps crashing.
Previously the M1 Mac mini used to crash but that *touch wood* has been stable since I removed Docker and Visual Studio Code (replaced the latter with the Insiders version). Hope I donāt jinx that next!
I used to love macOS ā still do ā but this is getting frustrating. :( Hope itās just a beta issue and is fixed in the subsequent releases.
Just to eliminate the NVMe I also booted with the SSD and tried. That crashed in a few hours! So at least itās not the NVMe.
As an aside I used AmorphousDiskMark to compare the disk performance between NVMe and SSD on the iMac. The former connects via the ASUS ROG enclosure I blogged about earlier, using Thunderbolt, the latter via regular USB 3.0.
Hereās iMac NVMe:

And iMac SSD:

About half the speed of NVMe, but thatās fine⦠in terms of perception I donāt think I actually notice it with day to day use. Sure, things like iCloud and Photos download faster with the NVMe; but thatās just a one time thing. I prefer the NVMe over the SSD mainly because it has 2TB vs 1TB and I can keep my entire iCloud downloaded on the NVMe.
Hereās stats from the Mac mini M1 though. :) Blows everything else away!

Hah. š Not sure itās all great though. Random 4k writes seem to be poor on the M1 compared to iMac NVMe or SSD.
Update (Friday; 26th Feb): Iāve stopped using the Magic Mouse to see if that makes a difference. I donāt think it will⦠but Bluetooth has been a bit flaky with the M1 Macs (not that the iMac is an M1) and maybe Apple is fiddling with the Bluetooth stack in Big Sur and each time the issue happened I was clicking on something ⦠and I am running out of ideas š so letās see how that goes. I am still using Bluetooth for my headset and I have a Magic Trackpad 2 connected, so this is not really a good test.
Update (Monday; 1st Mar): That didnāt help. For reference hereās a snippet of what I am getting:
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
panic(cpu 1 caller 0xffffff800efc18c6): Kernel trap at 0xffffff800f557fe4, type 14=page fault, registers: CR0: 0x000000008001003b, CR2: 0x0000700006d84180, CR3: 0x0000000022f42000, CR4: 0x00000000003626e0 RAX: 0xffffff7fa81674a0, RBX: 0xffffff869c89a4a0, RCX: 0xffffff800fe3e1e8, RDX: 0x0000000000000002 RSP: 0xffffffb0ac703d00, RBP: 0xffffffb0ac703d40, RSI: 0x0000000001e101e1, RDI: 0xffffff9363c9a100 R8:Ā Ā 0x000000000000ffff, R9:Ā Ā 0x00000000ffffffff, R10: 0x00000000000001e1, R11: 0xffffff9364f2a0c0 R12: 0x00000000dc000001, R13: 0xffffff9363c9a100, R14: 0x0000000000000000, R15: 0xffffff93652dafd0 RFL: 0x0000000000010297, RIP: 0xffffff800f557fe4, CS:Ā Ā 0x0000000000000008, SS:Ā Ā 0x0000000000000000 Fault CR2: 0xffffff7fa816753c, Error code: 0x0000000000000000, Fault CPU: 0x2, PL: 0, VF: 10 Ā <snip> Ā Process name corresponding to current thread: kernel_task Ā <snip> Ā last started kext at 20035428933: >!AActuatorDriver 4440.3 (addr 0xffffff800ffc1000, size 28672) last stopped kext at 141842220512802: @plugin.IOAVBDiscoveryPlugin 940.2 (addr 0xffffff7fa82d3000, size 32768) |
The 14=page fault seems to suggest RAM according to some forum posts but I am not sure thatās the case. Iāve always seen this AActuatorDriver as the last kext in previous crashes too. Part of me thinks this must be to do with the external SSD or NVMe as thatās the only unusual setup I am doing, but I donāt want to believe that either as Iād hate going back to using the iMac with its internal HDD.
The process is kernel_task so not any program as such. I donāt have TimeMachine backups so itās not that at least. I quit two more backgrounds programs anyways ā HazeOver, and Toothfairy ā to minimize the number of variables. Since then Iāve had no crashes but 2 out of the 3 crash free days were weekends and I donāt use the iMac much over a weekend, so maybe thatās why.
Today the Sculpt Keyboard froze as I was typing and I had a panic moment that the iMac was going to crash again. Luckily it didnāt but that made me also remove the Sculpt Keyboard dongle and switch to the Magic Keyboard. So thatās three changes Iāve made now and if things donāt crash for a while Iāll start undoing these to figure out the culprit.
Update #2 (Monday; 1st Mar): Crashed again! Phew. I was not too happy with the Magic Keyboard (having been used to the Sculpt keyboard for so long) so I was hoping it would crash lol!
Also, got a different message this time:
|
1 2 3 4 5 6 7 8 9 10 |
panic(cpu 2 caller 0xffffff800e35e9b8): watchdog timeout: no checkins from watchdogd in 92 seconds (23993 total checkins since monitoring last enabled) Ā <snip> Ā Process name corresponding to current thread: kernel_task Ā <snip> Ā last started kext at 226255085770152: >!AHIDKeyboard 224 (addr 0xffffff800c526000, size 16384) last stopped kext at 280424032980: >!AThunderboltEDMSink 5.0.3 (addr 0xffffff7fa42ba000, size 32768) |
Funny that this time the error is different and also mentions the keyboard. I think this oneās a side effect of me using the Magic Keyboard?
Also, interestingly this time I was on a Teams call even while the machine was stuck I could hear the audio (couldnāt see the video as I was on a different screen and couldnāt move back to the video screen). It stayed that way for about half a minute and then rebooted.
Update #3 (Monday; 1st Mar): No, it didnāt crash, but I figure I am using this blog post as a sort of notepad to myself on what I am doing so might as well blog it here. Since the iMac didnāt crash over the weekend I am thinking it must be something that gets triggered when itās active. Not sure what though.
I disabled an SSD connected to the system to see if that makes a difference. So as of now I have no HazeOver and ToothFairy, and no SSD. I donāt use HazeOver so I uninstalld it anyways. Will be installing Docker (again) now as I need it and had uninstalled in case that was a culprit.
Update (Thursday; 4th Mar): No crashes since the last update. Maybe the SSD was the culprit? Odd coz I was able to use it fine before switching to the NVMe, but now that I switched to the NVMe and the SSD is a clone of it maybe thereās some driver issue? Doesnāt make much sense. Could be ToothFairy too but unlikely as thatās not the process name. macOS 11.3 Beta 3 is out today but Iāll wait a day or two before jumping to that ā just to see if the current state is really stable.
Update (Saturday; 6th Mar): No crashes. Sweet! So it was the SSD for some reason (even though it worked fine for ages).
Update (Wednesday; 10th Mar): The machine went ahead and updated itself to Beta 3 on Monday. It hasnāt crashed like before but both yesterday and today when I tried to change the wallpaper to one of the default ones it crashed. Eugh! Both times it crashed with this error:
|
1 |
panic(cpu 1 caller 0xffffff800c34b9b8): watchdog timeout: no checkins from watchdogd in 97 seconds (16602 total checkins since monitoring last enabled) |
Ā
