I’ve been pushing my luck for a while running beta macOS software (started with Big Sur) and it’s finally catching up to me I guess. Funnily enough some months ago I had stopped using beta software and that’s 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 one’s been particularly worse for me. I run it on my iMac and it crashes every day (or at least every other day). I can’t find any crash logs but I get a message about the kernel and invalid addresses – sort of like this guy on the Twitters. I couldn’t find any other reports of this so maybe it’s just me and that guy. I don’t use Chrome though, only Firefox. I don’t think I have any crazy software installed either – just “standard” 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 don’t jinx that next!
I used to love macOS – still do – but this is getting frustrating. :( Hope it’s 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 it’s 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.
Here’s iMac NVMe:
And iMac SSD:
About half the speed of NVMe, but that’s fine… in terms of perception I don’t think I actually notice it with day to day use. Sure, things like iCloud and Photos download faster with the NVMe; but that’s 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.
Here’s stats from the Mac mini M1 though. :) Blows everything else away!
Hah. 😀 Not sure it’s all great though. Random 4k writes seem to be poor on the M1 compared to iMac NVMe or SSD.
Update (Friday; 26th Feb): I’ve stopped using the Magic Mouse to see if that makes a difference. I don’t 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 😀 so let’s 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 didn’t help. For reference here’s a snippet of what I am getting:
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: 0x000000000000ffff, R9: 0x00000000ffffffff, R10: 0x00000000000001e1, R11: 0xffffff9364f2a0c0
R12: 0x00000000dc000001, R13: 0xffffff9363c9a100, R14: 0x0000000000000000, R15: 0xffffff93652dafd0
RFL: 0x0000000000010297, RIP: 0xffffff800f557fe4, CS: 0x0000000000000008, SS: 0x0000000000000000
Fault CR2: 0xffffff7fa816753c, Error code: 0x0000000000000000, Fault CPU: 0x2, PL: 0, VF: 10
Process name corresponding to current thread: kernel_task
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)
14=page fault seems to suggest RAM according to some forum posts but I am not sure that’s the case. I’ve 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 that’s the only unusual setup I am doing, but I don’t want to believe that either as I’d hate going back to using the iMac with its internal HDD.
The process is
kernel_task so not any program as such. I don’t have TimeMachine backups so it’s not that at least. I quit two more backgrounds programs anyways – HazeOver, and Toothfairy – to minimize the number of variables. Since then I’ve had no crashes but 2 out of the 3 crash free days were weekends and I don’t use the iMac much over a weekend, so maybe that’s 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 didn’t but that made me also remove the Sculpt Keyboard dongle and switch to the Magic Keyboard. So that’s three changes I’ve made now and if things don’t crash for a while I’ll 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:
panic(cpu 2 caller 0xffffff800e35e9b8): watchdog timeout: no checkins from watchdogd in 92 seconds (23993 total checkins since monitoring last enabled)
Process name corresponding to current thread: kernel_task
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 one’s 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 (couldn’t see the video as I was on a different screen and couldn’t 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 didn’t 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 didn’t crash over the weekend I am thinking it must be something that gets triggered when it’s 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 don’t 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 there’s some driver issue? Doesn’t make much sense. Could be ToothFairy too but unlikely as that’s not the process name. macOS 11.3 Beta 3 is out today but I’ll wait a day or two before jumping to that – 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 hasn’t 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:
panic(cpu 1 caller 0xffffff800c34b9b8): watchdog timeout: no checkins from watchdogd in 97 seconds (16602 total checkins since monitoring last enabled)