Macs, Mavericks and audio issues, WARNING: LONG!

[Reezekeys]

Warning! If you're not using OSX Mavericks and/or not into extreme geekness, bail on this now! I won't be surprised or offended if this thread get zero views and replies!

 

The last few months have been a challenge for me and my normally ultra-reliable laptop rig running a rather involved setup of two instances of Bidule hosting a variety of plugins. I'll be on a gig and the audio will turn to fuzz or drop out for a few seconds, then come back. I believe, but am not positive, that this started with an OSX update (10.9 to 10.9.2). I've been all over Google and the Apple support forums, and saw one other person reporting my issue, but there is plenty of talk about somewhat similar audio problems with Mavericks. In my case, every time this happened, I would make a note of the time, then check my Console logs and see the same pair of messages repeated, sometimes over a thousand times:

 

kernel[0] IOAudioStream[0xffffff8026410c00]::clipIfNecessary() - Error: attempting to clip to a position more than one buffer ahead of last clip position (ba1,3f8)->(ba4,4f8).

 

followed by

 

kernel[0] IOAudioStream[0xffffff8026410c00]::clipIfNecessary() - adjusting clipped position to (ba4,3f8)

 

What makes this a real bitch to figure out is that it doesn't happen on every gig. And just to make it more interesting, I'll sometimes see a bunch of these messages in the log after a gig in which I heard no audio problems! (In those cases though, there are less of them than when the dropouts happen). This is one of those times that most clearly illustrates what differentiates us laptop guys from the regular hardware keyboard crowd. Without a little understanding of how computers work, you can really be up a creek here. I can handle a 2 second dropout at my local jazz jam gig but on a stage in front of a few thousand people, playing unaccompanied with a singer it's a different story (yes, it happened).

 

These messages are just telling me what I'm hearing. They don't tell me what's causing them to happen in the first place. Of course I looked elsewhere in the logs, trying to see if some OSX daemon started up at the same time, perhaps a process that suddenly sucked a lot of CPU juice and so might cause the audio issues. I didn't see anything obvious. I tried shutting down a few processes I thought might be starting up at times when I'd be on the bandstand: I signed out of iCloud (there are lots of system processes that "call home" to keep things synched). I disabled Notification Center (I once saw notifications appear on my screen right when the audio dropped out). Yes, a fishing expedition, and these didn't solve the problem.

 

The next idea was to create a new account and try running my music setup from there but that turned out to be a Pandora's box of reauthorizing software and trying to find & copy support files from my present account to the new one. I had the idea of rolling back to 10.9.0 but I'm not sure how I could do that (my laptop came with 10.9.0 but no install disks, and AFAIK you can't find 10.9.0 in the Mac App store anymore).

 

In an effort to not make this post longer than it already is, I'll just detail where I'm at right now: two days ago I updated to OSX 10.9.5 (I had gone to 10.9.4 but that didn't fix things). I spent a few hours at home, playing a looped sequence that gave my setup a workout, without issues. At my first gig with 10.9.5, last night, everything was peachy keen until 12:15AM when I heard the audio sputter for a second. This time I was ready: I had Console running, turned around and watched as a bunch of messages appeared. These were not the usual "ClipIfNecessary", they were "systemstats" and "spindump" processes writing a bunch of powerstat log files, all at once. I did a little research and found out that at 12:15AM every night, OSX is programmed to write out these log files, that they were probably done as part of the OSX development process and the engineers forgot to disable them, and that they could be safely nuked. This involved modifying a plist file or using a Terminal command (shall we talk about G7 #9 chords now? ). The semi-good news is that I actually have another gig tonight where I'll be playing past midnight, so I'll get a chance to see how this particular fishing expedition turns out!

 

If you've hung in this far, thanks! It's my hope that this thread, being in a forum with a few of us running OSX laptop rigs, might spur some dialog, or come up in a Google search by someone else dealing with this same issue. Maybe if we can find each other we can help each other! Now let's get back to those G7 #9s!

[mate stubb]

I've been avoiding Mavericks on my work computer for months. Some of our software dev tools have problems, and guys running 2 monitors report machines crashing regularly.

[Reezekeys]

Yea you'd think by the .5 update they could iron out some of these issues?

 

My fingers are crossed about the ClipIfNecessary problems. Hopefully 10.9.5 fixed that. In my case, I can't go back to a pre-Mavericks system: my laptop is a late 2013 MacBook Pro, it shipped with Mavericks.

[Markay]

Wonder if any MainStage 3 users have experienced this. MainStage seems to be very good at shutting down all background processes when in perform mode. There must be a global command or script floating around that does this. If it exists and you can find it you could run it before a performance.

[Six-string-man]

Every time I consider going the software route, I read something like this, and it keeps me on the straight and narrow.

 

I'm sorry for your troubles, but thanks for helping me to stick to hardware!

 

 

SSM

[Bill W]

Wonder if any MainStage 3 users have experienced this. MainStage seems to be very good at shutting down all background processes when in perform mode. There must be a global command or script floating around that does this. If it exists and you can find it you could run it before a performance.

 

I have not experienced any issues.

[Reezekeys]

if MainStage users are doing OK then it's probably something I installed or updated, or something got corrupted in my system. Reinstalling OSX is definitely one thing to try, although I'd like to go with the simpler (i.e., less time consuming!) alternatives first. The thing is, I'll do three or four gigs with no issues, think everything is fixed, then have it happen on the 5th gig. Yea, hardware guys, I'm a little jealous! But I have years of virtually trouble free operation with my previous two laptops, so hopefully this will just be a blip. And it's not like you guys are immune to breakdowns either!

[b3plyr]

No problems with MainStage 3, Pro Tools, etc., so far.

[WesG]

Sounds like you're short on IOs.

 

Is your disk an SSD? (It should be)

 

Is it possible you're swapping when these background processes kick off? You might want to max the RAM and configure the box for no swap during performances. Actually, you should probably also nuke whatever OS X uses in place of crond during a performance, too.

 

Clues?

- http://wiki.summercode.com/how_to_disable_or_enable_swapping_in_mac_os_x

- http://nathangrigg.net/2012/07/schedule-jobs-using-launchd/#launchcontrol

 

Wes

[Tusker]

Wonder if any MainStage 3 users have experienced this.

No problems here on Mavericks and Mainstage 3. Sorry to hear about this Reeze. Fingers crossed for you.

[PianoMan51]

Reeze, don't give up the faith! Don't go back to the Hw dark side...

 

Seriously, you got me turned on to Bidule, which I have used to heavily load my Macbook Pro running Mavericks 10.9.5. When you mentioned iCloud I jumped, because the only sure-fire way I know to get problems is to leave wifi on. For practice, no problem, but for gigs, never. Re: after-midnight gremlins, I rarely play that late and when I do the band is so effin' loud I might not hear a momentary dropout.

[Tusker]

The next idea was to create a new account and try running my music setup from there but that turned out to be a Pandora's box of reauthorizing software and trying to find & copy support files from my present account to the new one.

 

I believe this is still your best approach. Start clean again and climb back up the ladder of inference by adding plug-ins. Climbing down the ladder is never certain.

 

Good luck.

[Botanicus]

I've had brief dropouts with fuzz while working with Digital Performer 8.07, but also with iTunes, in 10.9.4. Upon immediate playback all will be smooth again, but it's annoying as hell.

 

I thought perhaps Time Machine was trying to back up at that time or something similar. I'll watch this thread with interest.

[Reezekeys]

Sounds like you're short on IOs.

 

Is your disk an SSD? (It should be)

My main drive, yes. For sample streaming I used a 128GB external SSD but switched to a USB3 flash drive. When the problem first started I immediately switched back to the SSD but the dropouts continued, so it wasn't the disk.

 

Is it possible you're swapping when these background processes kick off? You might want to max the RAM and configure the box for no swap during performances. Actually, you should probably also nuke whatever OS X uses in place of crond during a performance, too.

 

Clues?

- http://wiki.summercode.com/how_to_disable_or_enable_swapping_in_mac_os_x

- http://nathangrigg.net/2012/07/schedule-jobs-using-launchd/#launchcontrol

Thanks for those links. I just downloaded Launch Control, there's a lot to digest and it might be one or two steps beyond my pay grade but I'm hoping to find it useful. As far as the points you raise: no swapping is going on. I have Activity Monitor running and I monitor any swaps and my memory usage. I have 8GB of physical memory in my MacBook Pro, which is non-upgradeable. I'm always under that 8GB figure on my gigs, and "Swap Used" is always zero. I would expect this because an almost identical setup to what I use now ran for years on my previous laptop, a 2007-era MacBook Pro with Snow Leopard and 4GB, then 6GB of ram. No glitches on that machine, even at the same 128-sample buffer setting.

[Reezekeys]

Reeze, don't give up the faith! Don't go back to the Hw dark side...

 

Seriously, you got me turned on to Bidule, which I have used to heavily load my Macbook Pro running Mavericks 10.9.5. When you mentioned iCloud I jumped, because the only sure-fire way I know to get problems is to leave wifi on. For practice, no problem, but for gigs, never. Re: after-midnight gremlins, I rarely play that late and when I do the band is so effin' loud I might not hear a momentary dropout.

Yea Apple and iCloud means a very steady stream of communication between you and the mothership. I've been looking at Console logs a lot lately and there are a hell of a lot of calls to processes that try to phone home. Whether this can affect audio performance is an unknown. Turning off wifi doesn't necessarily stop these processes from running either.

 

As far as going back to hardware, I don't think so! I'm gonna get through this. What I have going in Bidule could not be done on any hardware workstation anyway.

[Reezekeys]

I believe this is still your best approach. Start clean again and climb back up the ladder of inference by adding plug-ins. Climbing down the ladder is never certain.

 

Good luck.

Thanks! Yea it might come to that, although I might try reinstalling the system first. AFAIK you can do that pretty easily on these newer machines without disturbing your user data.

[Reezekeys]

Just got back from the gig tonight. Zero audio issues and no "ClipIfNecessary" messages. I woke the laptop now (it's almost 1:15AM my time), don't see any spindump processes writing powerstat log files like last night either at least not yet. Fingers crossed as I have two bigger gigs coming up this weekend. Life on the edge!

[EscapeRocks]

2014 MacBook Pro 13". Mavericks. Mainstage 3. Arturia Prophet VS.

No issues.

[Reezekeys]

I found this via Google about four years ago someone at Metric Halo asked about the meaning of these "clipIfNecessary" messages. The answer, from an Apple engineer no less, is interesting:

 

http://osdir.com/ml/coreaudio-api/2010-08/msg00250.html

 

Quoting the relevant part: "The log message about this is probably unnecessary and should probably be relegated to debug builds."

 

I had seen this exchange before in my research, but now I started thinking that my audio issues might be caused just by these messages being written to Console. IOW, whatever is causing the messages to be written would not in themselves cause the audio to dropout it's just the reporting of them! So if there was a way to suppress these messages from being written, things might be made right again.

 

I can see that this thread is gonna be mostly me answering myself!

[keyman27]

I've got a MBA w/Mavericks. I've noticed a kind of 'fuzz' -- it's more like a broken speaker static sound. I thought there was something wrong with the hardware and was ready to visit the Genius Bar. It only happened with the onboard speakers, not the 1/8" line out or Thunderbolt-Firewire interface. Now I think it's only happening with Mainstage. And it comes and goes. It happened with one of the incremental updates to Mavericks.

[Reezekeys]

How often does this "fuzz" happen and for how long each time? Offhand it doesn't sound like the same thing I'm dealing with. Have you done the latest update to 10.9.5?

[OB Dave]

When the problem first started I immediately switched back to the SSD but the dropouts continued, so it wasn't the disk.

By that, do you mean you switched to the external SSD or the internal one? If you haven't already, you might try using only the internal drive to eliminate the possibility that USB is somehow hanging things up.

[Reezekeys]

When the problem first started I immediately switched back to the SSD but the dropouts continued, so it wasn't the disk.

By that, do you mean you switched to the external SSD or the internal one? If you haven't already, you might try using only the internal drive to eliminate the possibility that USB is somehow hanging things up.

External which I had been using without issues. But it was connected via the Thunderbolt-to-firewire adapter, so it wasn't using the USB bus (AFAIK anyway).

 

In the ongoing saga, I just examined my latest TM backup and it looks like Apple put a new version of the AudioAUUC.kext kernel extension with the 10.9.5 update. Of course I have no idea if that might have anything to do with what might be causing my issue but hey, it has "audio" in the title!

[Markay]

Hmm well due to the type of OSX install I have I know that kext's are used in OSX for device drivers. My guess is that Apple have added additional generic audio driver support that is misbehaving. I would log the issue with Apple support.

[Reezekeys]

My point was that the 10.9.5 update's new kext may have fixed things (I realize now how poorly it's worded, sorry!). One trouble-free gig does not prove that though. Stay tuned, as they say...

[Tusker]

Up to 10.9.5 here and not a problem ... Maybe I just skated past a bad OS update? Or maybe there are hardware/software contributing factors? :idk

 

Looking forward to hearing more Reeze ... Good luck.

[Ken Beaumont]

I recently had a similar issue fuzz when recording using my roland duo capture ex, lots of fuzz. I had not used it in about 3 months. Sure enough, there was a driver update because an apple update had broken the old one.

[Reezekeys]

Up to 10.9.5 here and not a problem ... Maybe I just skated past a bad OS update? Or maybe there are hardware/software contributing factors? :idk

Did you go right from 10.9.0 to 10.9.5? I did the .2 update which is where I think the dropouts started. Then I went to .4 and they continued. Now I'm at .5 with one trouble-free gig in the books. I can't be positive the .2 update broke things; I think I updated Kontakt too and I did make some minor changes to my Bidule layout. And considering the fact that sometimes I could go a few gigs with no issues makes this a real tough one to figure out. Nobody in the universe is gonna have my exact combination of hard & software.

[Reezekeys]

I recently had a similar issue fuzz when recording using my roland duo capture ex, lots of fuzz. I had not used it in about 3 months. Sure enough, there was a driver update because an apple update had broken the old one.

At least Roland got it together to update the driver quickly. I'm using the built-in audio but after the Apple debacle with the iOS 8.0.1 update, I wouldn't be surprised to learn they broke the driver they use for their built-in sound card!

[Reezekeys]

A new installment in the saga, for anyone interested...

 

OK, two nice gigs here in SoCal and zero issues with my rig. Rental companies and sustain pedals, that's another story but I'll save that rant for another time.

 

What I've been doing the last few gigs is running both Console and Activity Monitor with my two copies of Bidule. Console to watch the logs and AM to watch memory usage. After the set I sleep the computer, go back to the hotel room, wake it and look through the log files for anything unusual. What was unusual last night was that I did see the return of the "clipIfNecessary" messages, however they were written to the log at 10:30PM, as soon as I opened the lid to my MacBook Pro to wake it (the set I played was from 7 to 8:15PM). Strange because I'm fairly certain there was no audio going on when I woke the computer.

 

At exactly the same time as these new "clipIfNecessary" messages, the daemon "UserEventAgent" was called multiple times. This seems to happen every time the computer is woken from sleep. Googling has turned up issues with this process using excessive CPU in some instances. As I said, there was no audio being generated or passed when I woke my laptop but both copies of Bidule were still running so perhaps that had something to do with the clipIfNecessary messages that got spit out.

 

The bottom line, and a hopeful sign, is that during the gig itself there were no messages generated at all in the logs. Everything worked great. However these last two gigs were just one set, 65-70 minutes long, starting at 7PM. Not an entire night of playing like on my local gigs. I'm not pronouncing things fixed yet!