Hardware and performance

 
 
Picture of Mark Hardwick
System Locking up with more than a dozen users
 

I'm having trouble with my system locking up when more than a dozen people log on at one time to take a test.  By locking up I mean all computers freeze on a white screen and the icon just goes around and never loads the next page.  If some get a jump they can get logged in and start taking a quiz but when more people log on then all computers freeze.  We can eliminate users down to a dozen or less and everything is fine.   

This is happening at our classroom during a test and the internet is from a Net Gear router that is connected to a non-wireless Comcast box.  We’ve used this router for a couple of years now but just moved to Comcast from a DSL line and modem.  That’s when the trouble started.  I've had an IT guy look at it and does not see anything out of sorts.

I just want to know if there are any settings within Moodle that would limit the number of users at one time.  I'm fairly sure it's in the wireless system but just trying to cover my bases.  I don't have the option to run that many computers on hard wire so I can't really test it that way.

Moodle 2.2.4 build 20120706

Thanks for any input 

 
Average of ratings: -
Picture of Mark Hardwick
Re: System Locking up with more than a dozen users
 

Here's an update:

I changed the wireless router from an N600 to a 1750 Net Gear.  The problem is about the same when a dozen or so users tried to logon at the same time to do a quiz, the computers just lockup.  I'm thinking now that it's the Comcast service that’s the problem but not sure.  Today I hard wired 16-18 of the computer stations and I only have 8-10 on the wireless now.  It was a lot of trouble to do this just for diagnostic purposes because when the problem is fixed, I'll go back to wireless.  I'll find out tomorrow if the problem stays or goes away.  If the problem stays (which I think it will) then I guess it's a problem with the ISP.  It's just strange to me that we get good service on our office equipment and to mention again, we never had these issues with the old slower DSL service. 

A speed test shows about 20mbs down and 5mbs up whereas the old DSL line was 5mbs down and on a good day, maybe 2mbs up.  Is there a difference in stability and/or capacity between DSL and cable?  I did a speed test at a public place and it was .75 down and .25 up and the quizzes were slow but they kept going and worked.

Can you have fast internet service but fall short when it comes to capacity?  Like a race car and a one ton truck whereas it’s fast but just won’t carry the load?  I don’t really understand bandwidth and speed because I thought they went together.    I’m just trying to understand this.  I’ll repost the results of the hard wired system.

 
Average of ratings: -
Picture of Visvanath Ratnaweera
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers
Hi Mark

Just for the record: The other day a college had big problems with an online test, on Moodle 1.9. I showed them the monitoring graphs in the server (munin & Co.) to prove that the server is taxed not even 50% and the network doesn't even notice the load. They weren't convinced saying that they have invested a huge amount on money on latest desktops, fastest (20 Mbit/s down 10 Mbit/s up) network, network hardware and the very fast Windows 7 64 bit. The last one made me suspicious. I spend half a day there wrestling with their IT to get a second browser "just for comparison". Finally we got some "portable chrome" deployed and the problem solved! (Guess what the first browser was?)
sad
 
Average of ratings: -
Picture of Mark Hardwick
Re: System Locking up with more than a dozen users
 

Thanks for the reply.  I've been out for a few days and just now saw it.  That gives me something else to check and test.  IE on the browser??

The last run here at our location I changed as many as possible to hard wire and it was better but still a few hung up at random.  Some that locked up were wireless and some were hard wire.  Tomorrow I'm going to try all hard wire with a smaller class (but large enough to know if there's a problem) and I'm going to unplug the wireless router to make sure none of the computers are picking it up.  I wonder if there is a conflict between the two routers although the guy I use for IT when it gets over my head says it's set up right along with all of the networks.  Our hard wire network in the office is through the Comcast box/router and our wireless is fed from the Comcast box and only servers these lab computers for the students.  The hard wire network cannot see the wireless network and no one from the wireless can see my office system by design.

I feel pretty good about our host and his server because they log on and do quizzes remotly and no one has any trouble.  He also says none of his other customers are having trouble and he has checked and says he can't see anything from the server end.  I can't say they all log on at the same time in the evening but I've seen a large group logged on at several times.

I'll repost about how the hard wire test goes.   

 
Average of ratings: -
Picture of Visvanath Ratnaweera
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers
Yes, your first port of call in the case of client problems is to try alternative browsers. Well assuming your client machine is strong, but if it runs Windows it can't be weak.

Firefox and Firebug
http://docs.moodle.org/24/en/Performance_recommendations#Web_server_performance is the next. Check that whole document.

Since your network seems to be troublesome, you need to scrutinize that. On Unix/Linux there are plenty of built-in tools for that. I guess there are equivalent third party products for Windows.
 
Average of ratings: -
Picture of Ken Task
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers

Network comes before application and from what you've described (maybe not fully) suspect it's a combination of wireless + application.  It is possible to saturate a wireless base station.  Your work stations are using but what other/how many devices are also using?  Cell phones and/or tablets?  Have anything to monitor the status of the 1750 Net Gear?

Are there any other devices using wireless or radio frequencies that might possibly cause connectivity issues the base station desired?

Then there is the next hop ... base station to wire ... which I assume your Moodle server is wired.  Or is the Moodle server itself plugged into one of the wired ports on the router?

'spirit of sharing', Ken

 
Average of ratings: -
Picture of Justin Hunt
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers

I would be surprised if the problem was your network. Moodle is notoriously greedy with server resources, especially when running quizzes. Users running quizzes simultaneously and then finding the server locks them out is a fairly common occurrence here on the forums ... and I have battled with the problem a fair bit myself.

I know that different browsers perform differently but I don't think we can say Moodle tanks on IE and we should all use Chrome/Firefox. I think thats a footnote solution, and that the real solution lies in getting a server configuration that can handle the number of users Mark expects.

It sounds like you are on some sort of shared host Mark. So your options may be limited. The web server performance document Visvanath linked to on web server performance contains all the best advice, in the order of its effectiveness too. 

If you are on a VPS you *might* be able to get a php accelerator like APC going. That will give you immediate and good results. And tweaking your  Apache settings (if you are using Apache web server) might not speed things up much, but can help you prevent a total lockout situation. Information on both of those is in the document.

You may find however that a quicker path is to just upgrade to a hosting plan with more server resources. But failing that here are a few other footnote solutions that might help but that are not in the web server performance document:

* make sure you are running the cron job regularly. I had amazing slowdowns on one site when I had not set this up.

*  tell the users to login sequentially. So when the person on their right has logged in, then they can start. Anything you can do to stagger their access and avoid simultaneousness is good.

* avoid adaptive style quizzes where the users are constantly pressing the check  button to check their answer.

 

 

 
Average of ratings:Useful (1)
Tim at Lone Pine Koala Sanctuary
Re: System Locking up with more than a dozen users
Group DevelopersGroup Documentation writersGroup Particularly helpful Moodlers

1. We just did a load-test with 1,000 simultaneous quiz users in Moodle 2.4.x, and it was fine. This suggests that the caching changes in did in Moodle 2.4 (MDL-34399) bascially solved the long-standing problems with quiz performance. (If anyone has further similar data, please share it with me.)

2. So, if you care about quiz perforamance, you really ought to be running Moodle 2.4 or later, but acutally, if you care about your Moodle at all, you should running 2.4 or later, because that is the oldest supported version.

3. No reason to avoid adaptive style quizzes. The purpose of Moodle is to support learning, and they are great for learning. It is perfectly possible to run them at scale with an up-to-date Moodle install.

 
Average of ratings: -
You either love it or you hate it
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers

1000 simultaneous users! Are you sure. I for one have never seem any Moodle system be able to take that many users at one time
100mb per user is common place when undertaking quizzes. That means you would have to have at least 100GB of ram
We have two web servers with 4 cores and 12gb ram each and it will handle about 100 simultaneous users on a quiz

I know this isnt a solution to your problem but I am really not sure about those figures

 
Average of ratings: -
Tim at Lone Pine Koala Sanctuary
Re: System Locking up with more than a dozen users
Group DevelopersGroup Documentation writersGroup Particularly helpful Moodlers

Well, what is your definition of 'simultaneous users'?

What we simulated was 1000 users all starting a quiz within a few minutes of each other, then working through it with simulated auto-save every 1 or 2 minutes, then all submitting together after about 10 mins. It wasn't a perfect simulation, but since we were only expecting <100 users on a quiz in real use at the moment, it was more than enough to give us confidence in the system.

The hardware for this system is described here: https://moodle.org/mod/forum/discuss.php?d=237754#p1032551

 
Average of ratings: -
Picture of Justin Hunt
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers

Well it is a little hard to give advice without a little more information from Mark about his setup.  

I mean Tim's ability to handle 1000 simultaneous quizzers probably has a lot to do with a great hardware setup and the ability to configure all aspects of the Moodle installation.

I was assuming Mark is on a shared host, without the ability to do a whole lot more than configure his Moodle installation. On a shared host the installed PHP version can actually limit what version of Moodle you install also.

So perhaps Mark could tell us if it is shared hosting. If it is a VPS /standalone server what are the specifications? Is APC running? ...

(regarding adaptive quizzes, if 12 students hitting "start" on a quiz locks it up, 12 students hitting "check" on an adaptive question will lock it up too. Thats all. It is more about balancing a heavily loaded server than about adaptive questions.)

 
Average of ratings: -
Picture of Mark Hardwick
Re: System Locking up with more than a dozen users
 

I just double checked in: site admin > notifications and it says Moodle 2.2.4 build 20120706. 

Yes, we do use a web hosting company and the servers they use belong to them.  I don't know the specs on their server system but I know they host many sites that make us look like the little guy.  They recently upgraded their server and once we started having trouble they switched us over to the new server to see if that would help.  We tripped a few firewall parameters at first that they had bypassed for us on the old system and those shut us down completely for a few hours until they corrected them.  After adjusting a few firewall settings and white-listing our IP we didn't have any more trouble like that.

Today:  We had 15 users log on at the same time and do a quiz using adaptive mode.  The wireless router was unplugged and everything was hard wire.  We experienced some slowness for sure (click an answer, then check and sometimes it went through within a second or two and sometimes it took up to 10 seconds) but I don't think we had to do a hard shut down on any computer.  This was an improvement but still too slow.  

We use Respondus Lock Down Browser and I have re-down loaded their latest version once we started having trouble.  I can say without a doubt that an internet disruption during a quiz is un-recoverable and the computer must go through a hard shut down.  I found that having Google Chrome on the computers had a conflict with Respondus so it has been removed.  I know everyone does not share the same thoughts on the use of locking browsers and I know it’s not perfect but from a technical point it should work.

Our cable internet is 16 mbps down and 3 mbps up but we get 20 x 5 on a speed test.  Can 15 moodle quiz users bog down the speed so bad it intermittently drops off line or would it just get slower?  Their next package is 50 down and I think 10 up.  Would this solve the problem and allow us to use the wireless router again?

Thanks for all the comments.      

 
Average of ratings: -
Picture of Visvanath Ratnaweera
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers
Mark

You should get your basics straight. Right now you have tons of hypotheses, not a single one supported by any measurement. Instead you are trying or thinking of changing everything without knowing where the problem is.

Initially you said the server is clean. Now comes this story of a server (software?) upgrade and switching of servers (hardware?) because there was "something". And suddenly there is a firewall which gave (some?) trouble in the past!

The latest topic is upping the Internet connection from 20+5 Mbis/s to 50+10 Mbit/s. To support 15 quiz users?

Then, what is the connection between bandwidth to the Internet and your WLAN router? I mean, if the WLAN router is the problem, how could a faster Internet line solve that problem?

All this time you refuse to do a comparison against a standard browser (Forefox and Chrome). I understand that you want to have a lock down browser in the long run. Right now you looking for the cause of your problem!

A story, from this Sunday: A colleague said he is going to upgrade his Internet package, because it was "too slow". A few questions revealed the culprit: he has recently added a 22" screen to his micro-AT PC and viewing YouTube full-screen! The poor graphic card in the micro-PC was the cause of the problem.
 
Average of ratings: -
Picture of Justin Hunt
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers

As you can see from the replies to this thread, there really are a lot of things to consider. As Visvanath says, you should try a few tests to narrow down on the issue.

You seem pretty sure it is network related. So I would see if YouTube loads up and plays movies on 15 pcs at the same time ok. How about hammering a quiz on  http://school.demo.moodle.net ?  If you get similar slowdowns and lockouts, then that should tell you it is network (or browser) related. If not well that would indicate a server issue.

Justin

 

 

 

 
Average of ratings: -
Picture of Mark Hardwick
Re: System Locking up with more than a dozen users
 

Visvanath,

Much of the problem with the basics is that I don’t understand the topic of internet speed.  My background is not in IT and the biggest part of what I know about computers comes from the school of hard knocks.  

Here’s what I do know:

Our Moodle/Respondus system worked fine (but slow) on the old DSL service at 3-5 Mbps up and 1-2 Mbps down.

Our Moodle/Respondus system currently works fine at the hotel where most of our students stay during training at a whopping wireless speed of .75 Mbps down and .25 Mbps up.  The only thing is it’s hard to get 12 – 15 users on at the same time.  

When I changed to my ISP to a cable internet the problem started immediately.  This is a new service area for them and they just installed the lines and hardware.

We reused the same wireless router and password that we had for a year or two.

My host that I use had recently upgraded his servers to what he said would be way faster but he was switching his clients over a few at a time so he could work out any bugs.  I don’t know if it was a hardware or software upgrade or both.  My problem started while I was on the old server that worked, but he said if you’re having trouble I’ll switch you over to the new server.  We experienced a few firewall issues with too many wrong user name/passwords while introducing a new group of students to Moodle.  (We use the default password setting of a capitol letter, small letter and a character).  The new server firewall blocked our IP address and the hotels IP address the same day so he manually set these sites as OK/safe.

I replaced the router as previously mentioned but saw almost no improvement.  Replacing it was easier than hard wiring the classroom.

I moved some of these laptops to hard wire and the problem got better but still had locked ups.  We reduced the number to 10-12 and it works fine.

I moved all of them (15 in this class) to hard wire and they got through a practice quiz but took a while to load pages at times.  I’m sure 10-12 will work fine.

It would be a ton of trouble to load Chrome on all the laptops and try that.  I only get two chances a week to run these computers at our facility.  Not much time to mess with them and try new things.  I don’t understand how Respondus could work well with a few and not with a bunch.  I wouldn’t be surprised if another browser lightened the load and caused a group of computers ran better but that wouldn’t solve the problem.

The cable company has been out two times and says everything is ok.  They checked the incoming service and notified the area service team of a potential problem in the area.  I haven’t heard anything from them and doubt I will.

My IT guy has been out two times and can’t see anything.

The web host has checked the system more than once and can’t see anything.

I’m trying to understand internet speed and what I’m being told makes no sense.  I’m told that “bandwidth = speed” which I think is incorrect.  I understand they can be related but it seems to me that bandwidth is better defined as capacity which seems to be my problem because with a few computers running, the system works well.  On the other hand, I don’t understand why the 5 / 1.5 DSL worked better and certainly don’t understand why the .75 / .25 system works better.

Thanks,

Mark

 

Justin,

Just saw your post.  Yes we tried that about you tube and they played but at times they stalled out.

 
Average of ratings: -
Picture of Visvanath Ratnaweera
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers
Hi Mark

Only now I checked what "Respondus" http://respondus.com/products/respondus/ is. It is proprietary software: an authoring tool for quizzes and a lock down browser. Do you use both? I advice you to contact them for support, if you are going to stay with that browser.

Moodle has a great quiz engine, Free and Open Source. And there is the Safe Exam Browser http://safeexambrowser.org/news_en.html complementing it, based on Firefox, all F/OSS.
 
Average of ratings: -
Picture of Ken Task
Re: System Locking up with more than a dozen users
Group Particularly helpful Moodlers

Network comes before Applications.  So even if your Moodle server is just fine and doing all it can do, there still might be network issues that makes it *appear* to be a Moodle issue.  Although, many have started reporting slowness with Quizzes (considering what Moodle has to do to provide a quiz to students, it's somewhat understandable.  But, 'locking up', such you've described sounds more like in-complete rendering of pages in a browser - which shouldn't require a restart of the machine, but, rather, quitting the browser instance, and launching another.

In your summation, see 'ISP to a cable internet' and 'new service area for them and they just installed the lines and hardware'.

A quick explanation (poor one at that) about networking ... when a user uses a browser and request something from a web server ... any web server ... it does so by making 'request'.  Those request are not a single thing going across the connections, but multiple and each piece (packet) has a TTL - time to live - built into the packet.  If the packet arrives at the server and the server attempts to respond it ships out as fast as it can more packets.

If a packet of info can't reach it's destination within the time allotted in the TTL (time to live) the packets kill themselves ... obliterated ... blown to data bits ... gone ... killed on the wire.  Means server or station requesting info get/sends incomplete information as the packets responding to or requesting never arrive.

Bad routing can cause slowness and behaviors such as you are describing.

One way to test if it's a network issue is to do multiple traceroutes from both ends ... workstation to server ... and from server to workstation.

Both PC's and Mac's have command line utilities to trace the route.  On a Mac having such issues, from it's terminal, try this command:

traceroute moodleservers.fullyqualified.domainname

Like traceroute www.google.com

On a PC, think the command is tracert

If the network isn't locked down completely one will see a display of the routers/switches from the Mac to the destination.  30 hops is normally the max.  If traceroute cannot reach destination within 30 hops, it will die.

On your server, it too should have traceroute.  And, your apache server does log workstation IP addresses accessing it.  From the server terminal using traceroute, trace the route back to one of those IP addresses.

So, as an example, apache logs show an IP of 66.25.86.6x accessing something Moodle.   So traceroute back to the work station would be:

traceroute 66.25.86.6x

That trace will probably NOT reach the workstation as they are usually protected from direct access via a firewall ... actually, the IP address you see might be a gateway IP address from which multiple student machines come.

Comparing those traces (workstation to server - server to workstation) sometimes give clues - like double hops and times that should be 30ms or less in the range of 900ms or higher.  Ok, ms = milliseconds.

If it turns out to be a routing issue ... you cannot, by yourself, or any tweaking to your Moodle server ... fix it.  It's up to your provider.  They too might have to contact other networks to get the routing optimal.

For info on traceroute: http://bit.ly/1bcHwkj

'spirit of sharing', Ken

 
Average of ratings:Useful (1)
Picture of Mark Hardwick
Re: System Locking up with more than a dozen users
 

Visvanath,

We only use the lock down browser.  We used the authoring tool when we converted our question bank from our previous quiz system into Moodle a few years ago.  It allowed us to copy our questions, paste them into the tool and then import them into Moodle.  I'm sure an experienced person in Moodle could do it another way but at the time Moodle was new to us and it worked well on a one time basis.  It would be a good tool for an organization that has instructors that writes questions for a course but does not have administrative rights into Moodle.  Basically, write your questions and then bring them into Moodle.

Ken,

That makes sense.  I’m going to read it several more times to understand better about the flow and routes of information but when you mentioned that it could be killed if it took too long to make the trip, that sounds like what is happening.  The ISP is sending a tech out either tomorrow or Friday so I will press them a bit to look at the routing.  I’ll check with my IT guy and web host to see if they can put a trace on the routing and see what’s up.  It definitely has something to do with load/capacity because it works ok with a few so maybe the routing changes with a load or just slows down and loses it under load.

Thanks,

Mark

 
Average of ratings: -