ScreenConnect on CentOS is sluggish

[JaredBusch]

WARNING|CRITICAL logs (lasts 6 entries)
  2017-02-06 10:52:17 > 2017-02-06 10:52:24 CPU IOwait (62.7/66.0/69.3)
  2017-02-06 10:47:53 > 2017-02-06 10:48:00 CPU IOwait (60.9/64.1/67.3)
  2017-02-06 10:45:43 > 2017-02-06 10:45:46 CPU IOwait (63.5/63.5/63.5)
  2017-02-06 10:45:11 > 2017-02-06 10:45:14 CPU IOwait (60.1/60.1/60.1)
  2017-02-06 10:44:28 > 2017-02-06 10:44:31 CPU IOwait (65.7/65.7/65.7)
~ 2017-02-06 10:43:50 > ___________________ MEM real (1.58G/1.61G/1.64G) - Top process: mono

IOwait happening oftenish

[JaredBusch]

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Starting a support chat to get things reported, we shall see where this goes.

/sigh

[thwr]

deleted: mistake

[gjacobse]

How many clients are you running? We currently have 546 running without to much issue.

Obivous questions:

• centOS updated?
• nGinix running?
• System resources?

As SAM is more on the OS side, I'd say this is more up his alley than mine. So much of this is still Japanese translated Greek...

[JaredBusch]

Support Tech said:

Ok, 6.1 has our improved video encoding routine.
You may need to assign more CPU to that.
Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

[gjacobse]

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Support Tech said:

Ok, 6.1 has our improved video encoding routine.
You may need to assign more CPU to that.
Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

Yes - They push Windows OS pretty hard. Which is 'okay' but dang it,.. we are on centOS and that is where we are staying.

[thwr]

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Support Tech said:
Ok, 6.1 has our improved video encoding routine.
You may need to assign more CPU to that.
Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

Really? I should definitely stop developing bus drivers which run perfectly fine on Mono. How could I even think about doing such a thing?

[JaredBusch]

Hmm just clicked to a different group of machines and iotop shows this

[scottalanmiller]

@gjacobse said in ScreenConnect on CentOS is sluggish:

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Support Tech said:

Ok, 6.1 has our improved video encoding routine.
You may need to assign more CPU to that.
Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

Yes - They push Windows OS pretty hard. Which is 'okay' but dang it,.. we are on centOS and that is where we are staying.

Cheaper to push more CPU and RAM at it on Linux. We could double what we have and still be cheaper than where it used to be on Windows. Maybe even quadruple it. And it wasn't that much faster on Windows.

[scottalanmiller]

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

While native .NET does have advantages over Mono, that statement alone doesn't make sense. That's only one piece of the puzzle.

[thwr]

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

While native .NET does have advantages over Mono, that statement alone doesn't make sense. That's only one piece of the puzzle.

Totally agree here. Mono does a pretty good job in most cases, even performance-wise.

[scottalanmiller]

@thwr said in ScreenConnect on CentOS is sluggish:

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

While native .NET does have advantages over Mono, that statement alone doesn't make sense. That's only one piece of the puzzle.

Totally agree here. Mono does a pretty good job in most cases, even performance-wise.

And why even use Mono? Native .NET is available. I wonder why they don't mention it?

https://www.microsoft.com/net/core#linuxredhat

[gjacobse]

Current htop sorted by memory:

[JaredBusch]

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@thwr said in ScreenConnect on CentOS is sluggish:

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

While native .NET does have advantages over Mono, that statement alone doesn't make sense. That's only one piece of the puzzle.

Totally agree here. Mono does a pretty good job in most cases, even performance-wise.

And why even use Mono? Native .NET is available. I wonder why they don't mention it?

https://www.microsoft.com/net/core#linuxredhat

That is quite new, and they probably have not spent time to change because they prefer windows.

[scottalanmiller]

@JaredBusch said in ScreenConnect on CentOS is sluggish:

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@thwr said in ScreenConnect on CentOS is sluggish:

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

While native .NET does have advantages over Mono, that statement alone doesn't make sense. That's only one piece of the puzzle.

Totally agree here. Mono does a pretty good job in most cases, even performance-wise.

And why even use Mono? Native .NET is available. I wonder why they don't mention it?

https://www.microsoft.com/net/core#linuxredhat

That is quite new, and they probably have not spent time to change because they prefer windows.

I'm not surprised that they've not put time into it, but just saying that it's slow on Linux because of Mono needs a qualifier like "and we've decided to use Mono for now" or "Native .NET isn't ready yet and lacks something we need". The statement that they make is sensible only because we know Mono is being used by them, but on its own, the statement is weird because Linux is a fully viable Microsoft .NET platform now and it's been a big deal.

I'm not upset that they haven't tested or ported yet, but they could present that better, I feel. And it suggests that moving to Windows for .NET isn't a long term thing as we already have it native on Linux now. So we just have to wait for them to start using it.

[JaredBusch]

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@JaredBusch said in ScreenConnect on CentOS is sluggish:

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@thwr said in ScreenConnect on CentOS is sluggish:

@scottalanmiller said in ScreenConnect on CentOS is sluggish:

@JaredBusch said in ScreenConnect on CentOS is sluggish:

Our server definitely does perform better on a Windows OS as we use .NET instead of Mono.

While native .NET does have advantages over Mono, that statement alone doesn't make sense. That's only one piece of the puzzle.

Totally agree here. Mono does a pretty good job in most cases, even performance-wise.

And why even use Mono? Native .NET is available. I wonder why they don't mention it?

https://www.microsoft.com/net/core#linuxredhat

That is quite new, and they probably have not spent time to change because they prefer windows.

I'm not surprised that they've not put time into it, but just saying that it's slow on Linux because of Mono needs a qualifier like "and we've decided to use Mono for now" or "Native .NET isn't ready yet and lacks something we need". The statement that they make is sensible only because we know Mono is being used by them, but on its own, the statement is weird because Linux is a fully viable Microsoft .NET platform now and it's been a big deal.

I'm not upset that they haven't tested or ported yet, but they could present that better, I feel. And it suggests that moving to Windows for .NET isn't a long term thing as we already have it native on Linux now. So we just have to wait for them to start using it.

Looks like they are testing it. I asked about it.

That was a subject of our meeting with development last week. They hit some kind of roadblock, but it's definitely being looked into.

[scottalanmiller]

@JaredBusch Cool, if they had it out on that in 6-9 months I'd be thrilled. Our performance has been okay with Mono, but the whole Mono-wrapper thing is an unnecessary bit of overhead. I'm guessing the Linux server will surge forward in speed once they get that working. Hopefully no major roadblocks.

[JaredBusch]

Alright, upped the memory to 4gb for testing and wiped the sessions.db file.

So far only a single high CPU warning right at boot time, so going to expect/accept that.

2017-02-06 11:41:29 CPU user (82.1/84.6/87.0)

[JaredBusch]

Session.db changed dramatically.

[root@bnasc ~]# ls -l /opt/screenconnect/App_Data/Session.*
-rw-r--r--. 1 root root    794624 Feb  6 11:41 /opt/screenconnect/App_Data/Session.db
-rw-r--r--. 1 root root 322928640 Feb  6 11:29 /opt/screenconnect/App_Data/Session.db.2017.02.06
-rw-r--r--. 1 root root     32768 Feb  6 11:50 /opt/screenconnect/App_Data/Session.db-shm
-rw-r--r--. 1 root root   4128272 Feb  6 11:50 /opt/screenconnect/App_Data/Session.db-wal

[gjacobse]

@JaredBusch

Question on that - Do you have the system do DB maintenance?