Re: [ovirt-users] heavy webadmin

On Mon, 2016-04-04 at 14:51 -0400, Alexander Wels wrote:
On Monday, April 04, 2016 02:34:33 PM you wrote:
It is a hosted engine setup with the host on a physical server. I wasn't aware that the engine used the entropy. What do you recommend to fix it?
Okay, if it is a 3.6 then you should be able edit the hosted engine VM in the UI itself. I haven't done any self hosted engine myself but there are some things you can't edit, not sure if the rng is one of them. Anyway if you can edit the hosted engine VM then open up the advanced options (bottom left button in popup), then click the random generator side tab, and check the random generator enabled button (if it is not checked already that is).
As far as I know that should be enought to have the random passed from the host to the VM and that should improve the entropy on the host engine. You might have to restarted the hosted engine VM for it to take effect.
If the above is not possible or doesn't work, I would go with yum install haveged, then chkconfig haveged on, service haveged start or if it is centos 7 then its systemctl haveged enable and systemctl haveged start.
Sent from my Verizon Wireless 4G LTE smartphone-------- Original message --------From: Alexander Wels <awels@redhat.com> Date: 04/04/2016 13:57 (GMT-05:00) To: biholcomb@l1049h.com Subject: Re: [ovirt-users] heavy webadmin On Monday, April 04, 2016 12:55:55 PM you wrote:
On Mon, 2016-04-04 at 08:41 -0400, Alexander Wels wrote:
On Sunday, April 03, 2016 03:30:29 PM Brett I. Holcomb wrote:
On Sun, 2016-04-03 at 21:20 +0200, Nicolas Ecarnot wrote:
Le 03/04/2016 17:13, Greg Sheremeta a écrit : > We have patches in review that should fix this in 3.6.5. > The > underlying > problem is a couple of JavaScript memory leaks.
God bless you!
Since 3.6.2 or 3.6.3, the web admin is getting close to unusable after some minutes (Firefox or Chrome, windows or Linux). Alex Wels helped me try to debug this, but though much time spent on this, we failed. In a way, I'm satisfied other people are also expressing the same frustration about this issue, if that can help to debug.
Thank you.
Thank you also. I'm on 3.6.4 just released version and it can take 5+ minutes to go from the welcome page to the admin login page and then the interface is slow and frequently throws exceptions.
Two things to check if going from the welcome page to the login page is THAT show.
1. Make 100% sure your DNS is setup correctly (or your /etc/hosts if that is what you use). If the engine cannot resolve itself, it will create lots of issues. 2. If you are running hosted engine, make sure you have enough entropy, the login page generates a couple of tokens using secure random, which eats away at your entropy budget heavily and since hosted engine is a VM it is possible that you don't have enough entropy. You can check your entropy level with this:
cat /proc/sys/kernel/random/entropy_avail
There are several options for solving the entropy problem if that is the case.
DNS is 200% working . All hosts are resolvable from any where on the network and host can see engine and vice versa. Entropy problems makes more sense. I'd forgotten about this being a VM and I've run into the issue generating certs on VMware guests. I have entropy_avail value of 159. Is that good or bad and if bad what do I need to do.
159 is horrible to be honest, that is most likely the cause of your problems. Anything below around 300 is bad.
I am assuming this is not a hosted engine since you said you forgot this is actually a virtual machine. So you are running some other kind of VM manager to host the engine. The best thing to do is check if there is a way for your manager to pass sources of random to the VM. In oVirt its a couple of check boxes in the cluster setup and the VM setup. Obviously I don't know what you are using so I can't comment on that.
Another thing you can do is install some kind of psuedo random generator like haveged or rngd. I know for a fact you can simply 'yum install haveged' and it will work (someone else had the same problem and solved it that way).
It all depends on your level of needed security and VM manager, there are also physical sources of random generation like usb sticks and stuff.
I installed haveged and the available is up around 2500+ now. The server has a mouse and keyboard but it's not used much since connections are done via ssh. Thanks.

On Monday, April 04, 2016 10:00:45 PM Brett I. Holcomb wrote:
On Mon, 2016-04-04 at 14:51 -0400, Alexander Wels wrote:
On Monday, April 04, 2016 02:34:33 PM you wrote:
It is a hosted engine setup with the host on a physical server. I wasn't aware that the engine used the entropy. What do you recommend to fix it?
Okay, if it is a 3.6 then you should be able edit the hosted engine VM in the UI itself. I haven't done any self hosted engine myself but there are some things you can't edit, not sure if the rng is one of them. Anyway if you can edit the hosted engine VM then open up the advanced options (bottom left button in popup), then click the random generator side tab, and check the random generator enabled button (if it is not checked already that is).
As far as I know that should be enought to have the random passed from the host to the VM and that should improve the entropy on the host engine. You might have to restarted the hosted engine VM for it to take effect.
If the above is not possible or doesn't work, I would go with yum install haveged, then chkconfig haveged on, service haveged start or if it is centos 7 then its systemctl haveged enable and systemctl haveged start.
Sent from my Verizon Wireless 4G LTE smartphone-------- Original message --------From: Alexander Wels <awels@redhat.com> Date: 04/04/2016 13:57 (GMT-05:00) To: biholcomb@l1049h.com Subject: Re: [ovirt-users] heavy webadmin
On Monday, April 04, 2016 12:55:55 PM you wrote:
On Mon, 2016-04-04 at 08:41 -0400, Alexander Wels wrote:
On Sunday, April 03, 2016 03:30:29 PM Brett I. Holcomb wrote:
On Sun, 2016-04-03 at 21:20 +0200, Nicolas Ecarnot wrote: > Le 03/04/2016 17:13, Greg Sheremeta a écrit : > > We have patches in review that should fix this in 3.6.5. > > The > > underlying > > problem is a couple of JavaScript memory leaks. > > God bless you! > > Since 3.6.2 or 3.6.3, the web admin is getting close to > unusable > after > some minutes (Firefox or Chrome, windows or Linux). > Alex Wels helped me try to debug this, but though much time > spent > on > this, we failed. > In a way, I'm satisfied other people are also expressing > the > same > frustration about this issue, if that can help to debug. > > Thank you.
Thank you also. I'm on 3.6.4 just released version and it can take 5+ minutes to go from the welcome page to the admin login page and then the interface is slow and frequently throws exceptions.
Two things to check if going from the welcome page to the login page is THAT show.
1. Make 100% sure your DNS is setup correctly (or your /etc/hosts if that is what you use). If the engine cannot resolve itself, it will create lots of issues. 2. If you are running hosted engine, make sure you have enough entropy, the login page generates a couple of tokens using secure random, which eats away at your entropy budget heavily and since hosted engine is a VM it is possible that you don't have enough entropy. You can check your entropy level with this:
cat /proc/sys/kernel/random/entropy_avail
There are several options for solving the entropy problem if that is the case.
DNS is 200% working . All hosts are resolvable from any where on the network and host can see engine and vice versa. Entropy problems makes more sense. I'd forgotten about this being a VM and I've run into the issue generating certs on VMware guests. I have entropy_avail value of 159. Is that good or bad and if bad what do I need to do.
159 is horrible to be honest, that is most likely the cause of your problems. Anything below around 300 is bad.
I am assuming this is not a hosted engine since you said you forgot this is actually a virtual machine. So you are running some other kind of VM manager to host the engine. The best thing to do is check if there is a way for your manager to pass sources of random to the VM. In oVirt its a couple of check boxes in the cluster setup and the VM setup. Obviously I don't know what you are using so I can't comment on that.
Another thing you can do is install some kind of psuedo random generator like haveged or rngd. I know for a fact you can simply 'yum install haveged' and it will work (someone else had the same problem and solved it that way).
It all depends on your level of needed security and VM manager, there are also physical sources of random generation like usb sticks and stuff.
I installed haveged and the available is up around 2500+ now. The server has a mouse and keyboard but it's not used much since connections are done via ssh. Thanks.
Your login into the webadmin should be significantly improved now. No more 5 minute login times.

Le 05/04/2016 14:30, Alexander Wels a écrit :
2. If you are running hosted engine, make sure you have enough entropy, the login page generates a couple of tokens using secure random, which eats away at your entropy budget heavily and since hosted engine is a VM it is possible that you don't have enough entropy. You can check your entropy level with this:
cat /proc/sys/kernel/random/entropy_avail
159 is horrible to be honest, that is most likely the cause of your problems. Anything below around 300 is bad.
Alex, I checked that on my engines, and the entropy is around 120. But those are NOT hosted. Can you confirm this lack of entropy enduces no slowness on dedicated engines? -- Nicolas ECARNOT

On Tuesday, April 05, 2016 03:58:37 PM Nicolas Ecarnot wrote:
Le 05/04/2016 14:30, Alexander Wels a écrit :
> 2. If you are running hosted engine, make sure you have enough > entropy, the > login page generates a couple of tokens using secure random, > which > eats away > at your entropy budget heavily and since hosted engine is a VM > it is > possible > that you don't have enough entropy. You can check your entropy > level > with > this: > > cat /proc/sys/kernel/random/entropy_avail
159 is horrible to be honest, that is most likely the cause of your problems. Anything below around 300 is bad.
Alex,
I checked that on my engines, and the entropy is around 120. But those are NOT hosted.
Can you confirm this lack of entropy enduces no slowness on dedicated engines?
Lack of entropy is what causes super slow login. I creates some tokens on login that use up entropy. However once you have logged in it re-uses the token and the entropy should not be an issue after that for the most part. He didn't actually say but I suspect that after the initial minutes wait for the login, it was not THAT bad after that. Not like what we are seeing with your system.

On Tuesday, April 05, 2016 03:58:37 PM Nicolas Ecarnot wrote:
Le 05/04/2016 14:30, Alexander Wels a écrit :
> > 2. If you are running hosted engine, make sure you have > > enough > > entropy, the > > login page generates a couple of tokens using secure > > random, > > which > > eats away > > at your entropy budget heavily and since hosted engine > > is a VM > > it is > > possible > > that you don't have enough entropy. You can check your > > entropy > > level > > with > > this: > > > > cat /proc/sys/kernel/random/entropy_avail
159 is horrible to be honest, that is most likely the cause of your problems. Anything below around 300 is bad.
Alex,
I checked that on my engines, and the entropy is around 120. But those are NOT hosted.
Can you confirm this lack of entropy enduces no slowness on dedicated engines?
Lack of entropy is what causes super slow login. I creates some tokens on login that use up entropy. However once you have logged in it re-uses the token and the entropy should not be an issue after that for the most part.
He didn't actually say but I suspect that after the initial minutes wait for the login, it was not THAT bad after that. Not like what we are seeing with your system. _______________________________________________ Users mailing list Users@ovirt.org For me once I was on the admin login page things went much faster. Now
On Tue, 2016-04-05 at 10:04 -0400, Alexander Wels wrote: that I have haveged on the Engine VM the change from the welcome page to the admin login page is almost instantaneous.
participants (3)
-
Alexander Wels
-
Brett I. Holcomb
-
Nicolas Ecarnot