1

Hi All,

I have a concern over my free edition of UEB (I have registered on their forums, but for some reason it was a pointless exercise, i am not allowed to post or reply to posts - so i am posting my query here in the hope someone from Unitrends is able to see and assist me here.

The whole reason i got UEB was to backup my live machines on my lab ESXi system, only 2 are important to me but i do backup 4 (my license limit) - so far no problems, everything works well, i get nightly reports and very little if any failures (at least not from the backup - i have made a whoops once or twice though).

My System

I have 3 HP Microservers N40L 16GB (I know they are not greatly powerful but they suit my needs)

2 Netgear Readynas Ultra 6

TP-Link 24 port Gigabit switch

Each Micro has 4 ports connected and the NAS both have 2 connected

My backup method;

Incremental forever - total to backup is 205gb

My issue

Is not bandwidth, is not space, is not memory… It is constant CPU from UEB

I first thought this was because a job was running in the background, but it has been almost 100% CPU for nearly 2 weeks now. Now i know the AMD CPU in the Micros are DC 1.5Ghz and are no match for a real CPU but 100%… all the time - really?

When it is at 100% it is usually postgres: postgres bpdb that is eating it

I have it alloctaed it 2 vCPU so 3Ghz total

It would make sense if it was synthesizing, but it only shows as above and i don’t know why it is doing this almost all the time - perhaps i am missing something obvious here.

I am happy to switch to other backup methods if they will stop the high cpu, however i would like incremental still if i can.

As i said above this is Lab/live, i am happy if i need to, stop using UEB and just go back to file level backup from the guest directly, that said i do like UEB so far, i just would rather backup my systems without killing resources to do so.

Any ideas?

12 Spice ups
2

Hi

Can you check whether you have DeDupe turned on?

As a trial - turn it off for a few days - (and then check whether your CPU remains pegged)

Regards

PS: Do you have one of the N40L’s dedicated to running UEB? -

Or does UEB co-exist in one of the N40Ls with the other VM’s it might be backing up

1 Spice up
3

Each of the N40Ls are ESXi hosts, UEB is not dedicated to, but does end up on it’s own host (DRS puts it on it’s own host, so in reality yes it ends up being the only machine on the N40L). FWIW my other machines, even the live ones are usually less than 100Mhz of cpu time used - on average.

Native DeDupe is on - i shall turn it off and see how it does. I do see a ratio of 10-20% though - i guess i can live with more space used vs a flatlined CPU. This also would allow me to run a couple more machines.

DeDupe is now off, i shall run it a few days and report back - hopefully this will be long enough and will solve the issue

The VMs UEB backs up, since it has been installed have never been on the same host.

1 Spice up
4

Rod, How to I optimize or perform maintenance to the UnitrendsDataBase http://support.unitrends.com/ikm/questions.php?questionid=170

1 Spice up
5

Thanks for the link - i have seen it before but i didnt think i would need to run it since it runs weekly - and i’ve only been using UEB about a month.

I’ll see how the removal of Dedupe goes then run the commands, i will report back soon

Thanks to all.

1 Spice up
6

FYI - running the scripts hasn’t helped, nothing was removed and reindexing took about 2 seconds, i am guessing my tables are clean - but thanks again for the link.

1 Spice up
7

Rod - Sorry to hear about your problems with UEB. I checked with our forum master and due to an overwhelming amount of spam in the past, each post has to be manually approved, so if you’ve tried to post but haven’t seen it yet, approval may be pending. Can you check the system requirements to make sure that you meet these? Also, can you PM me your contact details so that I can have a member of the forum team look into your posts?

1 Spice up
8

Rod, I have 2 x n36 micro servers and have been surprised what they can do with hyper v 2012 on them. I run ueb, a 2003 print server and windows 7 machine on the same box and ueb behaves ok there.

Dave

9

Hi Katie - I was half expecting a response from you - and i am grateful that you have replied.

Yes my server meets the spec - it’s hard not to these days (but as my machines are not heavy load, originally i had given the UEB server 1vCPU and 2Gb ram - it ran fine for a week or so, then my problems started - so i increased the vCPU and still they persisted - see below, but adding ram i doubt will cure my CPU issue, that said it is claiming to only be using 57% of that 2gb currently)

As Dave mentioned he has 2 N36 - they are dual core 1.3Ghz AMD Athlon II Neo and the N40 is dual core AMD Turion II Neo @ 1.5Ghz. There is not much between them, a few tweaks here and there a slightly higher EVC mode on ESX on the N40L but otherwise same hardware and not so different spec.

Turning off the DeDupe hasn’t made any difference (at least not yet - i am going to reboot the guest shortly). If cpu returns to full belt after my reboot then something is clearly going on with the UEB system.

I did change the ram from 4gb to 2gb - but memory isnt my issue, cpu is and i am unable to allocate more cpu cores since i only have 2 per box. I could add up to 16gb to UEB but as above memory doesnt seem to be the issue, so i dont see increasing it as a help. Although i will to humour those who belive the memory will lower my cpu time.

@Dave6056 - They are not mighty beasts but they do hold their own and when running virtual machines HyperV, HyperV3 or ESXi they do pretty well. Not to mention their very low power usage and heat - this is the biggest factor for me

I currently run;

an RDP/VPN windows 8 machine used to connect to work

Kerio mail connect (email server), serv-u ftp, sawmill log reporting and spiceworks (all on one guest)

a W2012 Domain controller

Vcentre/vcenter server

Solarwinds virtual manager (trial)

UEB

Typically all bar UEB run on a single host but DRS wakes up my other host(s) because UEB is constantly glued to 90%+ of one hosts CPU.

All mentioned machines excluding UEB together only use about 10% cpu and 35% ram.

FYI - the reboot hasnt helped, it’s gone back up to 2.7Ghz

UEB did work fine for a few days, but has developed its issue in the last few weeks

@Katie I will PM you my email address shortly, i am out and about or at work a lot so my best contact method will be this.

You can see the screenshot - even though Dedupe is off it still has a little cpu time and postgres - well that is claiming more cpu time than i have.

UEB.jpg

10

Katie - i was going to ask

Would killing the machine and importing a new one from the download, adding my licence key and starting again - would this perhaps help?

I did have a little tinker at the start and may have pressed buttons and configs i wasn’t too sure about - although i did revert most so UEB is almost OOBE anyway - it can’t hurt right - just to speed up a fix?

11

Restarting it could make a difference. If you do this, I would suggest going with the NFR edition. It is free to all SpiceHeads and includes phone & email support. All you’ll need to do is enter in the code “SpicyUEB” on the registration page to get this version. Also, I would recommend watching the deployment video that (there is one for VMware & one for HyperV) is available on our website. It will walk you through all the steps to setup your UEB in the best configuration.

12

strace the PID of postgres: postgres bpdb and spam Katie so she can forward to support. LOL! Seriously though, if you have this issue (probably rare and isolated) so will someone else in the future at some point. The strace output should tell the dev guys something.

Cheers

13

Thanks all - and the Unitrends team, i really appriciate your help.

@Katie I had forgot about the SpicyUEB code - but 4 machines was good to get going, 1 question though (well maybe 2).

The 2 sockets protected is that total (if i have 3 hosts can i only protect 1, or can i protect mulitple hosts and multiple guests but no more than 2 physical CPU per host?) Based on the fact i am using ESXi 5 (update 2)

If it is the latter, unlimited hosts and guests, but a limit of 2 sockets and 2 physical machine - this would be handy for future VMs i build - i am not so worried about physical machines, i have these backup to one of my NAS and that replicates to the other periodically.

@boatsnjeepsnlinux - i have to say that i know the very very basics in linux - enough to get me through but for the majority it is all gibberish, i leave that kind of stuff the those who know it better, but thanks for the tip - who knows, i might get involved in it more some day.

A reboot didnt help, cpu is still very high, although it is no longer glued to 100% but it does ponder around 60-80 then spikes right back up again, it never seems to drop below the 60% mark (but i am not monitoring it all day, just when i spot it)

@64KB, i shall mail you shortly

Either way, if this is a rare case bug then i’m sure it will pop up again, and could be good for you guys to see why/how it is triggered, if it turns out to be me and my stupidity - lesson learnt

14

Rod - The NFR version will protect a total of 2 sockets, with unlimited clients from those two sockets.It will only protect 2 applications servers period, not 2 per host, and these can be physical or virtual, so if you don’t have a need for the physical protection, use these on a virtual client (these can be from sockets not protected by the 2 socket licensing).

15

Thank you for clarifying, i shall look in to changing my license from the free 4 machines to this version, possibly - lets see how 64KB gets on with my challenge before i make any other changes :o)

16

Hey Rod - did you ever get this resolved? I am seeing the same issue…

17

Hi Rod,

I am seeing the same issue too. I have even given the Unitrends VM 4 vCPU’s and 8GB of RAM… No luck. postgres keeps it slammed all the time. I too disabled dedup…

Thanks.

18

BrantWalsh/Rsnapp, apologies for the delay in replying - for some reason I never got an email alerting me to the first reply.

I can tell you that I never did get this resolved. I do understand that the HP Microserver’s CPU is by no means a monster, but I find it hard to believe the CPU would be bottle-necked continually for days if not weeks after any backup has completed, with no additional scheduled to take place.

I did however trial a complete new install, I set 1 server to back up once (note the once), no incremental, no additional full in the hope there would be a low CPU and no synthesizing - but I was wrong, even before the backups actually took place UEB consumed a whole core - adding a second only made it eat faster - but doing what I’ll never know.

I wasn’t going to spend too long investigating an issue that still seems software related to me.

Since then I have played with various other VM backup products - some of which should be ashamed of themselves. (I try to use free software or limited for home/test use to keep down costs) - after all I am only using these tools in my study lab setup.

What I will do for you guys over the next few days is another test. I no longer have any microserver’s assigned to ESXi, in place I have some HP ML310e Gen8’s (the microservers have been repurposed) For giggles I will set a backup with UEB on one of these and see how it goes - starting with what should be more than adequate for a single server backup of 2vCPU and 4GB ram.

I shall post back in a few days, but I have a feeling I will be answering as Rsnapp has.

I think adding more resources to UEB at the moment is like adding more fuel to an already roaring fire.

On the flip side I do like UEB’s interface - Web front end is certainly the way to go - unlike some other vendors I have tested still using applications - with no client-side connection (Really, this should be looked at).

For comparison, my current backup product is not in backup with anything and as I would expect the CPU is idle - using a massive 69Mhz - a more reasonable and expected figure for something that is not processing. It also runs flawlessly on 2vCPU and 2Gb ram.

I hope this helps for the time being.

Something else I noticed was under the free license you are able to download UEB updates, but under the NFR one you are not. So perhaps this is now fixed (an update resolves this perhaps) but NFR users might not be able to get the updates - I certainly wasn’t it said not allowed under my license.

Anyway, here goes - test number…

19

Ok, I am 1 day in to this and I have done as follows;

Installed a new UEB server (at present it is running on the trial with no license installed), I setup a backup location, left on dedupe and all OOTB features, set up a single machine to backup, incremental forever and I am seeing what I wanted to see previously (low CPU when not in use), with no backups running I am currently using 104Mhz.

I will be adding the original machines as mentioned in my first post back in to the mix - adding a little more workload. - The 4 original machines so I am replicating like-for-like (with the exception that I have new hosts now).

It will most likely be the weekend when I reply again, giving the system time to get a few backups behind its belt.

Out of curiosity though guys - what is the spec of the hosts you are using?

My original ESXi host was HP Microserver N40L dual core 1.5Ghz

My new host is rocking a quad core Xeon @ 3.5Ghz

To make the comparison as fair as I can I set EUB as follows;

Old UEB 2 x vCPU = ~3Ghz / 2GB ram

New UEB 1 x vCPU = ~3.5Ghz / 2GB ram

Backing up the same machines as originally, to and from the same location as previous, at roughly the same time at night

At the moment it is performing as I expected it to the first time round - but all that said, the first time I used it, it did actually work fine for a week or so - I guess when the synthesized full backup was made, this was when my issues started to become apparent.

If you guys have any other questions or would like to know any further on my configuration I’ll be happy to reply

Retention set to 1 (min) and 7 (max)

20

Rod

To confirm, with the NFR version you qualify for full support including all updates. The NFR qualification has to be renewed each year but that’s it. If you were unable to get updates, did you get any error messages?