Upgrade from 1.5.7 to 1.5.8 issues
Chris Whiteley last edited by Sebastian Roth
I am having significant speed differences during imaging from 1.5.7 to 1.5.8. I noticed this when I tried some of the 1.5.7.X revisions and so I went back to 1.5.7. I tried the update and here was the speed difference:
1.5.7 - 10.54 GB/min
1.5.8 - 7.23 GB/min
Is there a reason why there is such a drastic change in speed?
bzImage version init version init hash partclone version speed 4.19.101 1.5.8 … 0.3.13 slow 4.19.101 1.5.7 … 0.3.13 slow 4.19.65 1.5.7 … 0.2.89 fast 4.19.101 1.5.8mod … 0.2.89 slow 4.19.101 1.5.7 … 0.2.89 fast … … … … …
@Chris-Whiteley @george1421 I really like George’s truth table as it’s the best way to figure out those kind of things. To have it always on top I updated George’s information and put it into a table in the initial post. Let’s try to keep it updated there.
@Chris-Whiteley For now you have a path using the 1.5.7 inits. I might have time later tonight to unpack and repack the inits to copy over partclone. Once that is done we’ll have you test with the 1.5.7 inits with the patched in 0.3.13 partclone. Testing that will tell us about the next steps.
Thank you for your help til now to give us a clear picture of where the problem isn’t.
@george1421 Thanks for the update! Let me know what you guys find and if I can be of anymore help.
just tested with the 1.5.8 init.xz and got the slowness again
Ok so what the developers (or myself) need to do is pluck out partclone from the 1.5.8 inits and install it in the 1.5.7 inits. This will point exactly to partclone.
The issue with 1.5.8 (not a problem just many things change) is that the system used to create the virtual hard drive (init.xz) was upgraded, this also upgraded a number of support modules inside the init.xz. So on the surface we don’t know if the latest version of gzip is doing this or partclone. At least in my head.
@george1421 just tested with the 1.5.8 init.xz and got the slowness again
bzimage - 4.19.101
init.xz - 9133326
@Chris-Whiteley well for bzImage that’s easy
file bzImagewill tell you the current version of bzImage.
For init.xz its not as easy but not hard either. If we use md5sum utility we can get the fingerprint of init.xz file.
This is for 1.5.7
md5sum init.xz 913326f3317b577be3cb65a7bf332afb init.xz
If you have the chance, since something is different, can you test with the init.xz from 1.5.8? Its best to do all in one day if possible.
@george1421 Just as fast as when I had 1.5.7 and it was working. Both tests from normal SSD and NVMe worked as I think they should. How do I now get an output of what versions of things I have so that you guys can figure out what you need to?
This last test proved to be faster this time
Just as fast, or faster than last time?
Also this is why I wanted you to test the same day with the same set of circumstances. Just in case there was something in network land that was different than last week.
Chris Whiteley last edited by Chris Whiteley
@george1421 This is correct so far. This last test proved to be faster this time…hmmm…This has also been the most stable as far as not going from ridiculous speeds down to what it actually should be. Now it starts where it normally rests at.
@Chris-Whiteley That was the 4.19.101 version.
But maybe now I’m confused. I’m trying to build a truth table and I thought we zeroed in on something
1.5.8:bzImage 4.19.101 1.5.8:init.xz == Slow (partclone 0.3.13)
1.5.8:bzImage 4.19.101 1.5.7:init.xz == Slow (partclone 0.3.13)
1.5.7:bzImage 4.19.65 1.5.7:init.xz == Fast (partclone 0.2.89)
1.5.8:bzImage 4.19.101 1.5.7:init.xz == ?? (partclone 0.2.89)
Chris Whiteley last edited by Chris Whiteley
@george1421 I am testing it now. I found the 1.5.8 binaries.
@Chris-Whiteley If you are testing the bzImage from 1.5.7 and 1.5.8 on different days, could you test the bzImage from 1.5.8 now? I’m just trying to rule out other variables, because what you’ve told us, should not be. I’m not saying it can’t happen, only its a bit strange in the same series (4.19.x)
@Sebastian-Roth I just tested with the old bzimage for 1.5.7 and the speed was much faster and what I am used to.
So that would leave us with an issue in the Linux kernel or FOG server.
-OR- something outside of FOG causing the delay.
@Chris-Whiteley So that would leave us with an issue in the Linux kernel or FOG server. Kernel is easier to test. Leave the inits from 1.5.7 in place and grab the kernel binaries (
bzImage*) from the archive to use. See if it’s back to speed like this.
@Sebastian-Roth it was the same as the new init
@Chris-Whiteley So what’s the outcome of testing George’s suggestion? Sorry for nagging. Take your time…