Separate names with a comma.
Discussion in 'Windows Guest OS Discussion' started by RodB, Feb 27, 2017.
Okay, thank you.
UPDATE: I just had an occurrence of the issue again. Same situation as previously (Visual Studio is open to a solution but not actively working in it. Notepad++ was open and being used to edit a 1.5MB file) but I also had the "RT Disk optimization" is disabled. So, still happening despite that configuration change.
I'm going to close Visual Studio just to see if that is indeed a factor...
Noted, thank you! Looking forward to the results with Visual Studio being closed.
Disabling RT disk optimization didn't make any noticeable difference. I continued to experience the issue.
(NOTE: I've been out for a while which is why I haven't updated.)
I cannot say definitely that Visual Studio is the cause. However, I can say that I have experienced the issue when Visual Studio is open but that I have not experienced it when Visual Studio is not open. So, it appears Visual Studio may be a factor but the sporadic nature of this issue make it difficult to confirm absolutely.
But, again, I have never experienced this on my Windows 10 physical PC whether Visual Studio is open or not. So it is some combination of running Windows 10 on Parallels and (apparently) with Visual Studio open.
I would have expected this to be a more prevalent issue since I believe use of Visual Studio is a common reason for using Parallels on a Mac. But maybe others aren't using Windows 10. Or maybe my supposition is not accurate.
I will add this: Yesterday, my Windows 10 updated to 1511 (it was finally approved by our IT). Since that upgrade, it feels like the issue has been reduced. Again, I cannot be absolutely sure because of the inability to know when the issue will appear. But it has happened only twice today and, previously, I would have expected to experience it much more than that in the amount of time I've been working. Previously, it was enough that it was affecting my ability to work. Now, at least at the moment, it is only an annoyance.
Or maybe I'm just getting used to it. I welcome your thoughts on what else we might do or what other information I can provide you.
Thank you for the troubleshooting details. I have no explanation of this behavior yet, I suggest we try to localize the situation further:
1. Disable McAfee and see if anything changes
2. Change VM's vCPU number: setting 3 vCPUs and checking, then setting 5 vCPUs
3. If Visual Studio project is stored on Shared Folders, move it inside the virtual disk and checking.
@Korky Plunger: I was wondering if we were to compare which apps are running on our VMs, if we might find a common point -- or, at least, maybe narrow down the likely culprits. Are you still experiencing this issue?
Interestingly enough, just a few hours ago I was getting annoyed by the same "need to double click in notepad" issue that you mentioned earlier.
Whenever this happens, I always have the the following running:
- Visual Studio 2013
- Bvckup 2 (backup util copying some files from guest->host every few minutes)
- Occasionally notepad++, but often not
- Occasionally Chrome, but often not
During the 'freezes' (when the whole VM feels like it's paused -- different from the random input issues), I've noticed that task manager even stops updating, so it seems like this is more than an input issue. I recorded technical data and sent it to Dmitry, and his team said that input wasn't being processed due to CPU being potentially high, so I reduced my VM cores from 8 to 7, but this did not alleviate the issue. Also, I've had these pauses happen when the guest Windows VM was completely idle, and when the CPU usage on my mac was very low (<30%)..
Something is definitely rotten in Denmark, but I'm unable to definitively narrow it down as a user. I think the Parallels dev team will have to provide some kind of instrumented build that can track response time.
Thanks - very interesting regarding the task manager. I just received an update of Parallels this morning -- v12.2.0 (41591)
This morning, I've been trying a different test:
The only app I have open is Notepad++ -- nothing else, no VS, no SSMS, no File Explorer. I also closed my Caffienated just in case. (NOTE: Caffeinated is a small app I run in the background that keeps my VM awake so it doesn't time out and force me to login all the time. Since it does some magic under the covers to "move" the mouse in order to accomplish this, I thought it might be a factor.)
I moved the large file to a local (to the VM) folder to eliminate questions of the shared file system.
McAfee is still running on the VM as disabling this is problematic. I can disable it, but it does not stay once the next policy update occurs, which is every few minutes and the policy's aren't available for me to alter.
I'm running a timer to see if I can determine a time-related pattern.
So far, I've noted that the "double-click needed" behavior seems to proceed the complete pause.
Lately I've also encountered periods where the guest Windows OS is severely dropping input, but the host OS is registering the input with no problem. This comes and goes periodically throughout the day, even when neither the host nor guest OS is under any load. I finally remembered to gather technical data when this happened just now (Report ID 171544073).
I captured this data immediately after I notice the Windows OS dropping keyboard input when typing into Visual Studio. The problem occurred in Notepad++ as well at the same time. I was able to quickly switch back and forth between the host and guest OSes, and the issue did not happen on the host OS but happened on the guest OS within Parallels. To demonstrate the issue, between 2:40:45pm and 2:40:50pm Boston time (immediately prior to gathering the technical data) I started typing a sentence into Notepad++ and around 50-60% of the keys I typed were not registering. At the time (according to task manager) there was near-zero load on the guest OS, and the load on the host MacOS was around 10-15% (Parallels Desktop was using 9.3% CPU on the host OS).
I have the same annoying issue with Windows 10 guest OS in full screen mode. The mouse freezing is constantly occurs (approx 5 times in a day).
Even if VS2015 is not opened.
And even if just only chrome browser opened.
Fixed only by swithching between guest OS apps by pressing option-tab multiple times.
I'm experiencing another issue which I doubt it is related but I'd like to eliminate it as a consideration. Can you confirm that your browsers (IE, Edge) are working fine on your system? On mine, either browser fails to connect to any number of URLs, though they successfully connect to www.microsoft.com most of the time (scary).
IE is working fine for me.
Thank you. I was just getting on to report that I had resolved that issue. So, that behavior is no longer a consideration. Didn't think it really was anyway.
Thank you, KorkyPlunger, for the detailed issues description! I have reviewed the logs and it's full with the messages:
04-11 14:15:49.144 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.147 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.148 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.150 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.152 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.154 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.156 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.158 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
04-11 14:15:49.160 W /LocalDevices:36812:85007/ [CPs2Mouse] Mouse buffer full (guest hang?)
When it happens once again, please send us a spindump of the VM process (process name is the same as VM name).
Also, to work around this you can free another vCPU core to host system to allow Parallels Desktop better virtualize real-time devices (USB, sound and disk I/O) operations properly for the stable guest OS performance (currently 7 of 8 cores are assigned to VM).
Will be waiting for the spindump, and thank you in advance!
My Win10 VM is only configured for 4 cores the same as my Win7 VM. Below are other hardware configuration comparisons. NOTE: that RT Virtual Disk Optimization is turned off of the Win10 VM per your prior recommendation. I assume, since it made no difference, that this could be turned back on but I have not yet done so.
VM: Win10 / Win7
CPU Cores: 4 / 4
Memory: 8 GB / 4 GB
Hypervisor: Parallels / Parallels
Virtualization: Yes / No
Graphics - Memory: 512 GB / 512 GB
Graphics - Resolution: Scaled / Scaled
Graphics - 3D Acceleration: DirectX10 / DirectX10
Mouse - Optimize for games: No / No
Mouse - Smooth Scrolling: Yes / Yes
Keyboard - Optimize for games: No / No
Network: Shared / Shared
Hard Disk - Type: Expanding / Expanding
Hard Disk - Real-time Virtual Optimization: No / Yes
Thank you for the information! Can you also send us a technical data report after the next lockup, I will take a look at the logs.
Just sent report 172067707. Not sure if I captured the conditions or not as I fumbled getting the menu up to generate the report and the generation took quite a while. So, the condition may have ended by the time I got the report generated. The time of the event was about 10:27am.
When I reduce the number of vCPUs to 6, the pauses are shorter in time (making it difficult to have enough time to grab the spindump).
However, I'm not able to actually get a spindump. When I use activity monitor to generate a spindump, it completes immediately and just shows me a blank window. This occurs whether I try to take a spindump of my vm process or the parallels process itself. However, other apps on the system can be spindumped with no issue.
I'm able to use "sample process", just not "spindump".
Any idea what I'm doing wrong?
Also, I filed another report for a short VM pause I encountered today between 8:09am and 8:11am Boston time. Report ID 172273886
Thank you, Rod! But I couldn't find any clues in this one. Looks like the VM was idle 30 mins before you've send the report.
Please send me one more on the next occurrence, hopefully the comparison will help investigation.
Well, something is definitely wrong. It was certainly not idle at all. I was actively using the VM for a few hours before the event and when I sent that log. I will try to send another log on next occurrence, but the information you provided does not give me confidence given the log info is completely inaccurate.