I have a MFT pipeline as below:
FaceDraw and FaceDetect are custom MFTs. This pipeline is running fine for first few frames and all of sudden it exits.
I tried analyzing it using MFTrace and narrowed to a spurious evenet as below:
CMFPresentationClockDetours::GetTime #0000004BF1114E50 Time 20687020hns
This is not expected and suddently popping up. This event further leading to below set of events that are causing pipe to drain-off
CMFClockStateSinkDetours::OnClockStop #0000004BF98AFE88 System time 668039142ms
CMFClockStateSinkDetours::OnClockStop #0000004BF98AFB28 System time 668039142ms
CMFTransformDetours::ProcessMessage #0000004BF1838970 Message type=0x00000000 MFT_MESSAGE_COMMAND_FLUSH, param=00000000
So the question is, what are the ways to figure out the initiator for this event? Are there any better tools for debugging issues like this?
Initially I suspected EVR, but the removal of EVR from the pipeline is not making any difference.
Related
(Resolved: The delay/issues I was having, was not due to the delay of adding in try-catch, or redirecting errors to null. It was a result of using usernames that were not real, for testing the script, as opposed to actual/production names. The AD lookup times are obviously different -- didn't think of that before.)
I've been editing https://github.com/dafthack/DomainPasswordSpray for a work project, with the intent of adding in logging. The script outputs to a logfile about 6 times a second, which throws ioexception errors about accessibility of the file. It's a nonterminating error.
I'm ok with the error and potentially non-logged info. However, it clutters the console.
Since it's a non-terminating error, I tried set-content -erroraction 'silentlycontinue', but the error is still displayed. However, there is no measurable performance impact -- just a cluttered console.
So researching further, I tried a try/catch block, which does eliminate the error from the console, BUT the script now takes a tad over 2x as log to run.
Is there a 3rd alternative?
I can mitigate it by splitting the data lists into two, and running the script twice, concurrently. But it's less than idea, as I'm actually already splitting the data list up to speed up the process. (single list takes 2 hours, and 30m is the goal, so I'd need 8 windows to maintain current timing...)
Anyway. Hope that makes sense. Any thoughts/input appreciated. (attempting to copy the code to a machine where I can upload a portion here for those who want to review, but gmail blocks it. working on it.)
The code causing issues:
$FileLocation2 = "LOG-LastTested_$($Userlist)"
# Write out the last user tried
$tm = get-date
# This file is written to so rapidly, errors can occur, b/c it's still open, "In use by another process. SilentlyContinue not working, using try-catch which surpresses error (but is 2x slower).
Try {Write-output "Num: $cu, Name: $name, Pass: $Password, Time: $tm" | set-content -erroraction 'stop' $FileLocation2}
Catch [System.IO.IOException] {continue}```
Resolved: The delay/issues I was having, was not due to the delay of adding in try-catch, or redirecting errors to null. It was a result of using usernames that were not real, for testing the script, as opposed to actual/production names. The AD lookup times are obviously different -- didn't think of that before.
I was testing the error suppression (try-catch/redirect to null) with a list of dummy users. When I checked the rate of testing, it had dropped from 10 users/second with production accounts, to 3/s with dummy accounts. After reverting all the error suppression it was still at the slower 3/s rate. So I put my try-catch lines back in, and tried with production accounts, and it worked great -- 8-10 users/sec.
So, the issue was due to the only change: using dummy accts instead of production. Once I went back to all production accounts, WITH my error surpression -- worked like a champ. Hope that helps someone else.
Thanks to #RetiredGeek and #Mr.Sven who kept me testing and looking for a solution.
The message that is posted every time an execution fails is too verbose and creates too much noise. Is there a way to disable or hide it somehow? We have our own error messages within the scripts and don't need a red chunk of extra text displayed in the logs. Adding an error handler that will exit with code 0 is not an option because we still need the job to fail if a step fails.
That message is Rundeck standard output in case of failure (you can manipulate the loglevel but that doesn't affect the NonZeroResultCode last line, just the text from your commands/scripts), you can suggest that here. The rest are just workarounds as you mentioned.
This occurs even using the Quiet output.
I'm trying to run this (script.exs):
System.cmd("zsh", ["-c" "com.spotify.Client"])
IO.puts("done.")
Spotify opens, but "done." never shows up on the screen. I also tried:
System.cmd("zsh", ["-c" "nohup com.spotify.Client &"])
IO.puts("done.")
My script only halts when I close the spotify window. Is it possible to run commands without waiting for it to end?
One should not spawn system tasks in a blind hope they would work properly. If the system task crashes, some actions should be taken in the calling OTP process, otherwise sooner or later it’ll crash in production and nobody would know what had happened and why.
There are many possible scenarios, I would go with Task.start_link/1 (assuming the calling process traps exits,) or with Task.async/1 accompanied by an explicit Task.await/1 somewhere in the supervision tree.
If despite everything explained above you don’t care about robustness, use Kernel.spawn/1 like below
pid = spawn(System, :cmd, ~w|zsh -c com.spotify.Client|)
# Process.monitor(pid) # yet give it a chance to handle errors
IO.puts("done.")
I have a few questions regarding quartz.net.
What is it that keeps track of if there has been a missfire situation i Quartz.net?
What happens in the following scenarios:
If a job is run but cannot finnish due to some bug, does that count as a missfire or not?
What happens if i republish the solution, is the tracking reset?
Is there a way to receive information on what the scheduler has done and not been able to do?
I have the following code in my Run method:
IJobDetail dailyUserMailJob = new JobDetailImpl("DailyUserMailJob", null, typeof(Jobs.TestJob));
ITrigger trigger = TriggerBuilder.Create()
.WithIdentity("trigger1", "group1")
.WithCronSchedule("0 0 4 1 * ?", x => x.WithMisfireHandlingInstructionFireAndProceed())
.Build();
this.Scheduler.ScheduleJob(dailyUserMailJob, trigger);
this.Scheduler.Start();
The job is supposed to run the first every month on 4 am.
When testing I have set the system clock so that the jobb is missed for one month. According to the documentation when using WithMisfireHandlingInstructionFireAndProceed the job should be run the first thing that happens, but it dosent. Is there something wrong with the code or could it be some other reason the job is not run when using WithMisfireHandlingInstructionFireAndProceed() ?
If a job is missed, there is logic to bring it back. However, there is a "window" on how far back to go.
<add key="quartz.jobStore.misfireThreshold" value="60000"/>
You can increase this value.
If you have an ADOStore, misfires are persisted. Thus "if the power goes out", when restarting...you can recover from misfires.
If you have a RamStore...if "the power goes out", everything was in memory to begin with..so you won't get mis-fire handling, because everything was "in memory" and the memory is lost.
..
If you use Sql Server (AdoStore) and put a Profiler/Trace on it, you'll see the engine "poll" for misfires.......with a "go back this far in time" based on the misfireThreshold.
See this link:
http://nurkiewicz.blogspot.com/2012/04/quartz-scheduler-misfire-instructions.html
for more detailed info. Which has a "withMisfireHandlingInstructionFireAndProceed" note.
I am not sure what is going on, but i get this weird issue with buildbot.
The SVNPoller is configured as it should (checked various config example files), when i run the buildbot checkconfig it says that everything is fine....but it won't work at all.
If i trigger a build via the scheduler class it works fine, i can retrieve the source updates and build without problems (tried with a 1h timeframe).
The problem thou is that the poller is not working, so even if i build each hour, the changes column stays empty (i get the changes for the various versions thou, so if i click on the build detail i can see the sourcestamp carrying the right and most recent revision everytime that i modify the codebase); so I have no way to know if the build fails who did the last change.
Another peculiar thing is that in the twistd.log i see this line:
Warning: no ChangeSources specified in c['change_source']
And i am not sure why it wouldn't work since the checkconfig does not raise any error.
The result of this is of course that the only thing built is the hourly one, leaving me without the poller, and without knowing who is putting code in each build.
This is the code for the poller:
c['change source']=SVNPoller
(svnurl="svn+ssh://user#svnserver.domain.com/svn/project/trunk,
pollinterval=60*5,
histmax=10,
project=myproj,
svnbin = '/usr/bin/svn')
So far it looks good, so I am not really sure what is wrong here...why the SVNPoller is not triggering any build.
Anyone that has some suggestions about why is this happening ? Is there any other way to get changes from an SVN server? I am a total newbie at BuildBot and I am not really getting too much out of the manual; that looks much more like a scholastic book instead of being a manual that shows you how you do stuff :)
Thanks!!!!!
Ok, silly me :) the problem is the missing underscore on change_source...once added it the problem is solved
c['change_source'] = SVNPoller (svnurl=source_svn_url,
pollinterval=60,
histmax=10,
project='The_project',
svnbin= '/usr/bin/svn'
)
this will poll the svn codebase at source_svn_url (just put your svn:// path); and will check every minute to see if anyone has done changes; and will keep 10 changes in the record list (any change after the 10th will not show up so use it carefully if you do a lot of commits).
Hope that this helps who uses buildbot!