Slowdown Unicast and Multicast after upgrading FOG Server

mp12 · Feb 14, 2020, 3:59 AM

FOG-Server: VM with 6 cores and 8GB RAM
Ubuntu 16.04, FOG dev-branch 1.5.7.109 (before: 16.04, FOG dev-branch 1.5.3.?)

We are having a huge slowdown in deploying our images. Here is a comparison of Unicast and Multicast:

Unicast FOG 1.5.3: 21 minutes 20 seconds
Unicast FOG 1.5.7: 1 hour 7 minutes 40 seconds

Multicast FOG 1.5.3: 22 minutes 36 seconds
Multicast FOG 1.5.7: 3 hours 53 minutes 31 seconds

Image Size on Client around 250GB.
Compression Format: Partclone Zstd.
We just startet a new capture to see if the problem still exists.

Otherwise we have no clue what happened.

Anyone else having this problem?

george1421 · Feb 11, 2020, 11:44 AM

@mp12 said in Slowdown Unicast and Multicast after upgrading FOG Server:

Unicast FOG 1.5.3: 21 minutes 20 seconds
Unicast FOG 1.5.7: 1 hour 7 minutes 40 seconds

Translation to ~11GB/m for 1.5.3 and 3GB/m 1.5.7 (just to be clear we are talking about deployment speeds, right?)

I have to say I don’t believe you would get that change just upgrading from 1.5.3 to 1.5.7. Its not possible. The heavy workload is done by the target computer not the FOG server. Something else in your environment had to change. That is what my intuition is telling me.

With that said, It would be interesting to see what would happen if we used the older inits and kernel (from 1.5.3) in your current 1.5.7 environment.Those execute directly on the target computer, possibly impacting imaging speeds.

Sebastian Roth · Feb 11, 2020, 1:33 PM

@mp12 I would imagine this to be caused by the NVMe issues we have seen in the past weeks and months. Please edit one of your slow host’s settings and add nvme_core.default_ps_max_latency_us=0 as Host Kernel Arguments. Then try deploying again. If that doesn’t make a difference I have more things for you to try. Just give us a shout.

mp12 · Feb 11, 2020, 9:31 AM

Will quit work for now and restart at 8 a.m. CET

@Sebastian-Roth
No performance tweaks. Deployment around 1 hour.
Hardware Client: Dell Optiplex 9010, i7, 16GB RAM, SATA Samsung EVO 860 500GB

So we are ready to try the other things

@george1421
We tried different two older kernels but the results where even worse.

Sebastian Roth · Feb 11, 2020, 3:19 PM

@mp12 said in Slowdown Unicast and Multicast after upgrading FOG Server:

Hardware Client: Dell Optiplex 9010, i7, 16GB RAM, SATA Samsung EVO 860 500GB

Is this SAMSUNG 860 EVO M.2, 500 GB SSD? Yes, then create a post init script and add this command: nvme set-feature -f 0x0c -v=0 /dev/nvme0

If not we need to start looking at different things causing this.

mp12 · Feb 12, 2020, 7:13 AM

@Sebastian-Roth

It is a normal 2.5 in. SATA Samsung 860 EVO with 500GB (Model: MZ-76E500B/AM). Any other suggestions?

Sebastian Roth · Feb 12, 2020, 4:32 PM

@mp12 said in Slowdown Unicast and Multicast after upgrading FOG Server:

It is a normal 2.5 in. SATA Samsung 860 EVO with 500GB

Ok, I was totally on the wrong track with it being a NVMe issue.

Then we need to start looking at the components one by one to figure out what’s causing the slowdown. I would start with a single host as multicast testing is way more complex. Schedule a debug deploy task (just as if you create a normal task but make sure to check the box for debug before clicking create task button) for one host. Let it boot up to the terminal. To get the NFS share mounted you run the command fog and hit ENTER a few times till you see the message about share being mounted and checked. Then just hit Ctrl+c to stop the deploy and get back to the terminal.

Now we will run a few different tests:

mkdir /mnt/ramdisk
mount -t tmpfs -o size=1024m tmpfs /mnt/ramdisk
dd if=/images/#IMAGENAME#/d1p2.img of=/mnt/ramdisk/test.img

This will kind of test the network speed. It will copy about 1 GB of data from your FOG server to the client without writing to the local disk. Make sure you put in the correct #IMAGENAME# and depending on your disk layout you might need to choose a different partition file (d1p1.img or d1p3.img) which is larger than 1 GB to copy from.

Now we’ll dump that file from local RAM to disk. No network involved. Be aware this will wipe the data off your drive! You’d need to properly re-deploy this client after the test.

dd if=/mnt/ramdisk/test.img of=/dev/sda

Now take a picture of the screen and post that here in the forums!
For a clean shutdown you can run:

umount /mnt/ramdisk
umount /image
halt

This is mostly from the top of my head so don’t hesitate to ask if this doesn’t work as described.

mp12 · Feb 13, 2020, 8:21 AM

@Sebastian-Roth

Here the results:

I also copied some rows out of mysql fog.tasks. Here you can see the difference between speed and duration.

First a normal multicast with FOG 1.5.3.* (dev-branch). Sorry I don’t know the exact version anymore.

| 17919 | Multi-Cast Task | 2020-01-24 08:11:41 | 2020-01-24 08:12:33 | 143 | 389 | 4 | 0 | fog | 0 | 0000-00-00 00:00:00 | 8 | 0000000100 | 12.42GB | 00:21:00 | 00:00:00 | 249.006 GiB | 100 | 249.124 GiB | 1 |  1 |   | 0 | 1  |   |   |

This is the last single deploy I made with FOG 1.5.3.* (dev-branch):

| 18027 | Deploy Task | 2020-02-10 15:16:44 | 2020-02-10 15:17:34 | 214 | 391 | 4 | 0 | fog | 0 | 0000-00-00 00:00:00 | 1 | 0000000100 | 12.72GB | 00:20:38 | 00:00:00 | 250.320 GiB | 100 | 250.650 GiB | 1 | 9 |   | 0 | 1 |   |   |

Here a multicast after upgrading to FOG 1.5.7.109 (dev-branch):

| 18032 | Multi-Cast Task - 04-634-Reihe-1 | 2020-02-10 16:40:43 | 2020-02-10 16:42:14 | 214 | 391 | 4 | 0 | fog | 0 | 0000-00-00 00:00:00 | 8 | 0000000100 | 1.13GB | 03:52:47 | 00:00:00 | 250.422 GiB | 100 | 250.650 GiB | 1 | 1 |   |  0 | 1   |    |    |

And this single deploy was created with FOG 1.5.7.109 (dev-branch) the day after:

| 18042 | Deploy Task - 04-628-36 | 2020-02-11 12:53:24 | 2020-02-11 12:54:13 | 347 | 392 | 4 | 0 | fog | 0  | 0000-00-00 00:00:00 | 1 | 0000000100 | 4.61GB | 00:56:53 | 00:00:00 | 250.371 GiB | 100 | 250.620 GiB | 1 | 9 |  |  0 | 1 |   |    |

Quazz · Feb 13, 2020, 7:06 AM

@mp12 Those speeds are about 4 times as slow as we’d expect for a SATA SSD.

Can you also try the command hdparm -I /dev/sda in debug?

mp12 · Feb 13, 2020, 11:32 AM

@Quazz

Here the output:

/dev/sda:

ATA device, with non-removable media
	Model Number:       Samsung SSD 860 EVO 500GB               
	Serial Number:      S3Z2NB1KA50028H     
	Firmware Revision:  RVT01B6Q
	Transport:          Serial, ATA8-AST, SATA 1.0a, SATA II Extensions, SATA Rev 2.5, SATA Rev 2.6, SATA Rev 3.0
Standards:
	Used: unknown (minor revision code 0x005e) 
	Supported: 11 8 7 6 5 
	Likely used: 11
Configuration:
	Logical		max	current
	cylinders	16383	16383
	heads		16	16
	sectors/track	63	63
	--
	CHS current addressable sectors:    16514064
	LBA    user addressable sectors:   268435455
	LBA48  user addressable sectors:   976773168
	Logical  Sector size:                   512 bytes
	Physical Sector size:                   512 bytes
	Logical Sector-0 offset:                  0 bytes
	device size with M = 1024*1024:      476940 MBytes
	device size with M = 1000*1000:      500107 MBytes (500 GB)
	cache/buffer size  = unknown
	Form Factor: 2.5 inch
	Nominal Media Rotation Rate: Solid State Device
Capabilities:
	LBA, IORDY(can be disabled)
	Queue depth: 32
	Standby timer values: spec'd by Standard, no device specific minimum
	R/W multiple sector transfer: Max = 1	Current = 1
	DMA: mdma0 mdma1 mdma2 udma0 udma1 udma2 udma3 udma4 udma5 *udma6 
	     Cycle time: min=120ns recommended=120ns
	PIO: pio0 pio1 pio2 pio3 pio4 
	     Cycle time: no flow control=120ns  IORDY flow control=120ns
Commands/features:
	Enabled	Supported:
	   *	SMART feature set
	    	Security Mode feature set
	   *	Power Management feature set
	   *	Write cache
	   *	Look-ahead
	   *	Host Protected Area feature set
	   *	WRITE_BUFFER command
	   *	READ_BUFFER command
	   *	NOP cmd
	   *	DOWNLOAD_MICROCODE
	    	SET_MAX security extension
	   *	48-bit Address feature set
	   *	Device Configuration Overlay feature set
	   *	Mandatory FLUSH_CACHE
	   *	FLUSH_CACHE_EXT
	   *	SMART error logging
	   *	SMART self-test
	   *	General Purpose Logging feature set
	   *	WRITE_{DMA|MULTIPLE}_FUA_EXT
	   *	64-bit World wide name
	    	Write-Read-Verify feature set
	   *	WRITE_UNCORRECTABLE_EXT command
	   *	{READ,WRITE}_DMA_EXT_GPL commands
	   *	Segmented DOWNLOAD_MICROCODE
	   *	Gen1 signaling speed (1.5Gb/s)
	   *	Gen2 signaling speed (3.0Gb/s)
	   *	Gen3 signaling speed (6.0Gb/s)
	   *	Native Command Queueing (NCQ)
	   *	Phy event counters
	   *	READ_LOG_DMA_EXT equivalent to READ_LOG_EXT
	   *	DMA Setup Auto-Activate optimization
	    	Device-initiated interface power management
	   *	Asynchronous notification (eg. media change)
	   *	Software settings preservation
	    	Device Sleep (DEVSLP)
	   *	SMART Command Transport (SCT) feature set
	   *	SCT Write Same (AC2)
	   *	SCT Error Recovery Control (AC3)
	   *	SCT Features Control (AC4)
	   *	SCT Data Tables (AC5)
	   *	reserved 69[4]
	   *	DOWNLOAD MICROCODE DMA command
	   *	SET MAX SETPASSWORD/UNLOCK DMA commands
	   *	WRITE BUFFER DMA command
	   *	READ BUFFER DMA command
	   *	Data Set Management TRIM supported (limit 8 blocks)
	   *	Deterministic read ZEROs after TRIM
Security: 
	Master password revision code = 65534
		supported
	not	enabled
	not	locked
		frozen
	not	expired: security count
		supported: enhanced erase
	4min for SECURITY ERASE UNIT. 8min for ENHANCED SECURITY ERASE UNIT.
Logical Unit WWN Device Identifier: 5002538e408a5e55
	NAA		: 5
	IEEE OUI	: 002538
	Unique ID	: e408a5e55
Device Sleep:
	DEVSLP Exit Timeout (DETO): 50 ms (drive)
	Minimum DEVSLP Assertion Time (MDAT): 30 ms (drive)
Checksum: correct

george1421 · Feb 13, 2020, 6:48 AM

@mp12 I assume you ran the last hdparm from a debug console on the target computer. If so lets run this one too hdparm -Tt /dev/sda That should give us the disk performance test. I’m not totally convinced its a target computer issue, but we need to start collecting data where we can.

mp12 · Feb 13, 2020, 12:53 PM

@george1421

/dev/sda:
 Timing cached reads:   29692 MB in  1.99 seconds = 14911.18 MB/sec
 Timing buffered disk reads: 1614 MB in  3.00 seconds = 537.87 MB/sec

george1421 · Feb 13, 2020, 6:59 AM

@mp12 Read performance is what I would expect from an SSD drive.

The next bit we will test write performance. For this we need to collect the structure of the existing SSD drive. What we need to find is a partition that has at least 1GB of disk space.

Show me the output from this command: lsblk (executed on the target computer in a debug console)

NOTE: The document I’m working from is referenced here: https://forums.fogproject.org/topic/10459/can-you-make-fog-imaging-go-fast

Quazz · Feb 13, 2020, 7:04 AM

@george1421 He did some write tests earlier at around ~100MB/s from RAM to disk using dd.

george1421 · Feb 13, 2020, 1:07 PM

@Quazz Interesting, on the previous dd test. I would like to see this dd test and then the next step is an iperf test. That will test local disk and then network without involving the nfs stack or partclone. At least in my mind is how I would break it down. Something had to have changed besides fog.

mp12 · Feb 13, 2020, 7:56 AM

@george1421

Server

-----------------------------------------------------------
Server listening on 5201
-----------------------------------------------------------
Accepted connection from x.x.x.x, port 50672
[  5] local x.x.x.x port 5201 connected to x.x.x.x port 50674
[ ID] Interval           Transfer     Bandwidth
[  5]   0.00-1.00   sec   108 MBytes   903 Mbits/sec
[  5]   1.00-2.00   sec   112 MBytes   942 Mbits/sec
[  5]   2.00-3.00   sec   112 MBytes   942 Mbits/sec
[  5]   3.00-4.00   sec   112 MBytes   942 Mbits/sec
[  5]   4.00-5.00   sec   112 MBytes   942 Mbits/sec
[  5]   5.00-6.00   sec   112 MBytes   942 Mbits/sec
[  5]   6.00-7.00   sec   112 MBytes   942 Mbits/sec
[  5]   7.00-8.00   sec   112 MBytes   942 Mbits/sec
[  5]   8.00-9.00   sec   112 MBytes   942 Mbits/sec
[  5]   9.00-10.00  sec   112 MBytes   942 Mbits/sec
[  5]  10.00-10.04  sec  4.35 MBytes   937 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bandwidth       Retr
[  5]   0.00-10.04  sec  1.10 GBytes   939 Mbits/sec   11             sender
[  5]   0.00-10.04  sec  1.10 GBytes   938 Mbits/sec                  receiver
-----------------------------------------------------------
Server listening on 5201
-----------------------------------------------------------

Client

Connecting to host x.x.x.x, port 5201
[  5] local x.x.x.x port 50674 connected to x.x.x.x port 5201
[ ID] Interval           Transfer     Bitrate         Retr  Cwnd
[  5]   0.00-1.00   sec   113 MBytes   947 Mbits/sec    3    258 KBytes       
[  5]   1.00-2.00   sec   112 MBytes   943 Mbits/sec    0    364 KBytes       
[  5]   2.00-3.00   sec   112 MBytes   939 Mbits/sec    2    232 KBytes       
[  5]   3.00-4.00   sec   112 MBytes   943 Mbits/sec    1    318 KBytes       
[  5]   4.00-5.00   sec   112 MBytes   943 Mbits/sec    2    211 KBytes       
[  5]   5.00-6.00   sec   112 MBytes   943 Mbits/sec    0    364 KBytes       
[  5]   6.00-7.00   sec   112 MBytes   943 Mbits/sec    1    267 KBytes       
[  5]   7.00-8.00   sec   112 MBytes   943 Mbits/sec    1    364 KBytes       
[  5]   8.00-9.00   sec   112 MBytes   943 Mbits/sec    0    366 KBytes       
[  5]   9.00-10.00  sec   112 MBytes   943 Mbits/sec    1    282 KBytes       
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-10.00  sec  1.10 GBytes   943 Mbits/sec   11             sender
[  5]   0.00-10.00  sec  1.10 GBytes   942 Mbits/sec                  receiver

iperf Done.

Quazz · Feb 13, 2020, 1:33 PM

@mp12 I found somewhat similiar situations on google where the conclusion was to use Secure Erase on the drive first. Worth a try just to see if it helps?

mp12 · Feb 13, 2020, 3:11 PM

@Quazz

Did my best. After trying I noticed that the “frozen state” was active. Removing the SSD power cable while PC was powered on removed the frozen state. Then I was able to do a secure erase. A dd from ramdisk to /dev/sda was at the same speed as before. So no luck at all.

mp12 · Feb 14, 2020, 3:12 AM

@Sebastian-Roth @Quazz @george1421

I have some good and bad news.

First the good ones:

I created a deploy using the FOG.Wiki https://wiki.fogproject.org/wiki/index.php/Debug_Mode#Win_7
Therefor I booted a Clonezilla (2.6.4-10-amd64) Flashdrive and mounted the NFS-Share from FOG Server.

I started the deploy with the following command:

cat /images/IMAGEPATH/d1p2.img | zstd -d | sudo partclone.restore -O /dev/sda2 -N -f -i

Tried a deploy with the fog client and still one-third of the expected speed.
I think there is something wrong in the deploy process.
Main difference I can see is that the FOS uses Partclone 0.3.12 and Clonezilla 0.3.13.

Quazz · Feb 14, 2020, 3:44 AM

@mp12 I vaguely recalled someone having this problem before with similar outcomes.

Linking here for further reference: https://forums.fogproject.org/topic/13733/hp-elitebook-830-gen-6-issues-capturing-images-and-deploying-images/10

Our kernels and inits have since received a few upgrades, however.

Are you on the dev-branch, by the way? I don’t believe 1.5.7 was launched with partclone 0.3.12 (that’s for the upcoming release).

If not, then try the init and kernel files from https://dev.fogproject.org/blue/organizations/jenkins/fos/detail/master/122/artifacts

Slowdown Unicast and Multicast after upgrading FOG Server

152

12.0k

17.3k

155.2k