Page 1 of 2

vSRXNG problems

Posted: Thu Aug 24, 2017 4:34 pm
by andrewarp
Hello,

I've been playing around with eve for a while, and it is working kinda weird, i mean, most of the images that im using are Juniper ones. So, to be accurate on this topic, yesterday i was playing around with 2 vSRX15 (srxng called in eve), and it was going REALLY SLOW (im using a 64GB RAM Eve and 16 vCPU on eve), i mean it should be FAST, im hosting EVE OVA on VMware Workstation 12 on top of Ubuntu Desktop 16.04, the booting and the experience of the CLI of the SRX is going really really really slow, i tried using the default specs of the eve, and then i tried to upgrade those specs, default (2 vcpus, 4 ram) i tried 4 vcpu and 8 ram... and just the same... i dont get it, and then i tried to use 4 vsrx, and the first 2 crashed down, and the others didnt prompt anything... so, is there a bug with this version of srx? i havn't tried the vMX yet, so have anyone experienced this before?

If you need further information, just tell me and ill provide it to you.

Thanks in advance!

Re: vSRXNG problems

Posted: Fri Aug 25, 2017 2:21 am
by ecze
show us the output of
lscpu


from the host and from eve...

E.

Re: vSRXNG problems

Posted: Mon Aug 28, 2017 3:02 pm
by andrewarp
Eve,
https://ibb.co/kF2QnQ

and host,

https://ibb.co/fdE5nQ

i tried the vSRX on vmware and works like a charm ..

Re: vSRXNG problems

Posted: Tue Aug 29, 2017 9:41 am
by miklasz
hi,
I am using both OpenStack VM and Bare Metal,

openstack is much slower compare to bare metal, and bare metal is working beautiful, I am also investigating that "slowness" so if you get some more info will let you know, or if you figure it out let me know please

Bartosz

Re: vSRXNG problems

Posted: Fri Sep 15, 2017 3:38 pm
by packetferret
I am by no means an expert, but it looks as if there's room for optimization on the KVM emulation.

I've noted my qemu-system-x86 PIDs associated to the vSRXNG never go under 100% CPU utilization, even when it's running under CPU Limit = TRUE

Code: Select all

top - 10:37:24 up 37 min,  2 users,  load average: 1.00, 1.02, 0.97
Tasks: 259 total,   1 running, 258 sleeping,   0 stopped,   0 zombie
%Cpu(s): 12.1 us,  0.9 sy,  0.0 ni, 87.1 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem : 24661312 total, 18361604 free,  4031768 used,  2267940 buff/cache
KiB Swap: 25157628 total, 25157628 free,        0 used. 20106532 avail Mem 

   PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                                                
 17465 root      20   0 5083752 1.684g  21100 S 100.3  7.2   5:19.32 qemu-system-x86                                                                                                        
  7196 root      20   0 2930252 1.850g  20788 S   3.0  7.9   2:39.32 qemu-system-x86                                                                                                        
   161 root      25   5       0      0      0 S   0.7  0.0   0:59.73 uksmd                                                                                                                  
   550 root      20   0       0      0      0 S   0.3  0.0   0:00.79 kworker/2:2                                                                                                            
 14045 root      20   0   42212   4032   3304 R   0.3  0.0   0:01.49 top                  


Re: vSRXNG problems

Posted: Fri Sep 15, 2017 7:39 pm
by mike
There is no optimization fro KVM environment.
All tests and development is done on bare metal or ESXi environment.

There is no point reporting problems under KVM, since it won't be answered / taken care of in the near future.

Re: vSRXNG problems

Posted: Sat Sep 16, 2017 11:55 am
by Chris929
Of course it is slow.
im hosting EVE OVA on VMware Workstation 12 on top of Ubuntu Desktop 16.04
You run Ubuntu (first layer) --> Workstation(second layer) --> EVE OVA (third layer) --> vSRXNG (fourth layer)... This has to be slow - no matter what you optimize.

I tested all Juniper related images for multiple EVE-Versions on ESX (OVA) and Bare-Metal - and they all run amazingly fast - Bare-metal even faster than the ESX-Version since the last 2 or 3 Versions - I also blogged about this.

Why not convert your Ubuntu into EVE and skip the Workstation part? Every layer you add makes the images significantly slower - and the vSRX-NG or vQFX don't like to run in the layers 4 or higher - even in the third-layer they are already slower than you want them to be.

Re: vSRXNG problems

Posted: Wed Jan 03, 2018 1:14 pm
by lboue
Hello,

I have the same issue on a baremetal server with 64GB RAM. This is very slow with vsrxng-17.3R1.10 even with 8G RAM.

Code: Select all

top - 14:19:04 up 27 days,  4:15,  3 users,  load average: 7,49, 7,49, 7,29
Tasks:   1 total,   0 running,   1 sleeping,   0 stopped,   0 zombie
%Cpu(s): 55,0 us, 31,1 sy,  0,0 ni, 12,1 id,  1,7 wa,  0,0 hi,  0,2 si,  0,0 st
KiB Mem : 65961832 total, 10580324 free, 17798840 used, 37582668 buff/cache
KiB Swap: 67096572 total, 67096300 free,      272 used. 46439484 avail Mem 

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                        
10998 root      20   0 12,311g 3,316g  20780 S 199,3  5,3  39:31.63 qemu-system-x86       
This is better with the old version vsrxng-15.1X49-D80.4.

Code: Select all

top - 14:20:08 up 27 days,  4:16,  3 users,  load average: 7,26, 7,41, 7,27
Tasks:   1 total,   0 running,   1 sleeping,   0 stopped,   0 zombie
%Cpu(s): 16,3 us,  7,1 sy,  0,0 ni, 76,2 id,  0,4 wa,  0,0 hi,  0,1 si,  0,0 st
KiB Mem : 65961832 total, 10573432 free, 17805520 used, 37582880 buff/cache
KiB Swap: 67096572 total, 67096300 free,      272 used. 46432896 avail Mem 

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND                                                                                        
32601 root      20   0 5135176 2,940g  21048 S 140,0  4,7 440:51.74 qemu-system-x86
Any ideas?

Regards,
Ludovic

Re: vSRXNG problems

Posted: Mon Jan 08, 2018 1:10 pm
by Chris929
What CPU does your Bare-Metal-Server have?
In General the vSRX 2.0 is very resource-intense - if you want to just learn the concept of junipers SRX, then Firefly Perimeter will be way better.
Of course Juniper modifies KVM from time to time so check maybe the D100 or D110 - they run fine on my ESX-EVE.

Regards
Chris

Re: vSRXNG problems

Posted: Mon Jan 08, 2018 1:14 pm
by lboue
Hello Chris,

Here is the CPU we have for our baremetal EVE server: Intel® Xeon® Processor E5504 (4M Cache, 2.00 GHz).
You mean trying with 15.1X49-D100 version?

I tried last week with vsrxng-15.1X49-D80.4, it's a little bit better than 17.3.

Regards,
Ludovic