RFT: Kernels 2.6.27.4-linode14 and 2.6.27.4-x86_64-linode3

Linode Staff

Over the past few weeks, I've been messing around with 2.6.27.x based Xen guest kernels, and I'm pretty satisfied. These kernels use the new paravirt-ops interface, and will finally allow us to move away from the more-than-two-year-old 2.6.18 tree.

I'd appreciate some wider testing of these kernels – available now under your configuration profile's kernel drop-down.

The sourceballs are also located here:

http://linode.com/src/

Thanks!

-Chris

34 Replies

Just switched my dev linode over to this kernel. I'll let you know if I uncover any issues.

Installed and rebooted.

So far no problems. I will let you know how it does after a few days.

Wrong file extension: http://linode.com/src/2.6.27.4-linode14.tar.bz2

# file 2.6.27.4-linode14.tar.bz2 
2.6.27.4-linode14.tar.bz2: gzip compressed data, from Unix, last modified: Thu Nov  6 14:14:17 2008

@dfelicia:

Wrong file extension
Do'h… fixed, thanks.

@cburgess:

Installed and rebooted.

So far no problems. I will let you know how it does after a few days.

Ditto.

I'll try the new kernel at some point, but at the moment I want to see what kind of uptime I can achieve on my setup…

````
dfelicia@catch-22 ~ $ uname -a
Linux catch-22 2.6.27.4-linode14 #1 SMP Thu Nov 6 09:22:58 EST 2008 i686 Intel(R) Xeon(R) CPU L5420 @ 2.50GHz GenuineIntel GNU/Linux
dfelicia@catch-22 ~ $ uptime
20:58:47 up 2 days, 11:52, 1 user, load average: 0.12, 0.08, 0.01

````

So far so good. In fact, though it could quite possibly be due to some change/improvement by Internet provider, I swear access to my node is faster. When I ssh, I'm logged in instantly. My homepage also loads instantly (http://www.donsbox.com/). Previously, there was always a 1-2 second delay "waiting for donsbox.com…"

Any change from 2.8.18 -> 2.6.27 that could account for my now seemingly lower latency?

@dfelicia:

Any change from 2.8.18 -> 2.6.27 that could account for my now seemingly lower latency?
That's several years of kernel development… it could've been anything :)

@dfelicia:

Any change from 2.8.18 -> 2.6.27 that could account for my now seemingly lower latency?
Perhaps your use of a kernel from the future was affecting the space time continuum negatively thus increasing your latency.

MMMMM just started running it myself and I don't know if it's just a "Placebo" effect but my site feels a bit snappier also.

I'm running 32bit Gentoo with Drupal

http://www.rejecttheherd.net/

I'll let you know if I see any adverse affects

> Any change from 2.8.18 -> 2.6.27
OK, I made a typo. You guys are brutal :-)

BTW, like marcus, I'm running 32-bit Gentoo.

Something interesting I've noticed, my CPU % has drastically dropped since switching to the new kernel 8)

Two things I've noticed are that that bug in the CPU usage graphs were it was a perpetual ~3% is no longer there and I'm only seeing 2 CPU cores now instead of 4 as with 2.6.26.x.

@zengei:

I'm only seeing 2 CPU cores now instead of 4 as with 2.6.26.x.
Reboot and you'll see 4 cores again. This was a temporary misconfiguration on our end.

-Chris

@caker:

@zengei:

I'm only seeing 2 CPU cores now instead of 4 as with 2.6.26.x.
Reboot and you'll see 4 cores again. This was a temporary misconfiguration on our end.

-Chris
Great, thanks.

Just booted into 2.6.27.4-linode14 on my VPS (which serves out my blog {The Proliferation of Linux} and my site).

So far everything seems to work. I think that things might be a bit more zippy now, that's a purely subjective observation.

The server (a Linode 360) runs:
* *Wordpress MU

*Varnish 2

*Lighttpd (and PHP through FastCGI)</list></r>

@k33l0r:

So far everything seems to work. I think that things might be a bit more zippy now, that's a purely subjective observation.

Here's some results from Apache benchmark (quite impressive, I like to think):

hex-vps ~ # ab -n 10000 -c 50 http://proliferationoflinux.org/
This is ApacheBench, Version 2.3 <$Revision: 655654 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/

Benchmarking proliferationoflinux.org (be patient)

Server Software:        lighttpd/1.4.20
Server Hostname:        proliferationoflinux.org
Server Port:            80

Document Path:          /
Document Length:        29109 bytes

Concurrency Level:      50
Time taken for tests:   1.392 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      295349990 bytes
HTML transferred:       291090000 bytes
Requests per second:    7184.97 [#/sec] (mean)
Time per request:       6.959 [ms] (mean)
Time per request:       0.139 [ms] (mean, across all concurrent requests)
Transfer rate:          207234.49 [Kbytes/sec] received

I switched to 2.6.27.4-linode14 and rebooted, and everything came up perfect. Granted I don't do much more than serve up websites, ftp, and svn. But everything worked just great.

1) Munin's CPU graphs no longer showed idle time correctly.

2) Munin's display of memory cache looked most peculiar for my site.

3) Three times in a couple of days I had unusual 'disk io' alerts for no reason I could find.

I went back to the standard "latest 2.6" as before. Munin graphs now look normal. I'll wait and see about disk io alerts.

James

@zunzun:

I went back to the standard "latest 2.6" as before. Munin graphs now look normal.

Munin's memory cache graph now looks as it used to, fairly smooth, not anywhere nearly so chopped up and fragmented as with the newer kernel.

Link to munin image of cache memory showing this:

http://zunzun.com/choppy_cache.png

Link to munin image of CPU showing bad display of idle cpu:

http://zunzun.com/cpu_idle.png

James

Hmm. My munin graphs looks fine using the new 2.6.27: http://www.donsbox.com/munin/

What version of munin are you using? I'm on 1.34.

@dfelicia:

What version of munin are you using? I'm on 1.34.

from http://packages.ubuntu.com/intrepid/munin Ubuntu Intrepid is at 1.2.6, and according to the munin web site at http://munin.projects.linpro.no/ that is indeed the latest stable version. The version you are using must be from the far future.

I don't normally use this kind of arcane technical jargon, but my graphs are definitely poopy-doopy.

James

I am running Gentoo x64 and this new kernel is unable to boot properly.

Not sure why (tell me how to find out and I'll cooperate) but this bug applies even to default Gentoo installation.

 * Setting system clock using the hardware clock [UTC] ...
 * Cannot access the Hardware Clock via any known method.
Use the --debug option to see the details of our search for an access method.Cannot access the Hardware Clock via any known method.
Use the --debug option to see the details of our search for an access method.
 * Failed to set clock You will need to set the clock yourself
                                                                          [ !! ]
 * Configuring kernel parameters ...                                      [ ok ]
 * Updating environment ...                                               [ ok ]
 * Cleaning /var/lock, /var/run ...                                       [ ok ]
 * Wiping /tmp directory ...                                              [ ok ]
 * Device initiated services: udev-postmount
 * Setting hostname to fell ...                                           [ ok ]
 * Loading key mappings ...                                               [ ok ]
 * Setting terminal encoding to UTF-8 ...                                 [ ok ]
 * Setting user font ...                                                  [ ok ]
 * Starting lo
 *   Bringing up lo
 *     127.0.0.1/8
                                                                          [ ok ]
 *   Adding routes
 *     127.0.0.0/8 ...                                                    [ ok ]
 * Initializing random number generator ...                               [ ok ]
INIT: Entering runlevel: 3
 * Starting metalog ...                                                   [ ok ]
 * Starting eth0
 *   Bringing up eth0
 *     66.246.76.xxx
                                                                          [ ok ]
 *   Adding routes
 *     default via 66.246.76.1 ...                                        [ ok ]
 * Mounting network filesystems ...                                       [ ok ]
 * Starting local ...                                                     [ ok ]

After that, it just hangs up forever.

@drake127:

I am running Gentoo x64 and this new kernel is unable to boot properly.
It's booting all the way, you're just not getting a getty login prompt on the correct console device node.

Do you have "Xenify" set to Yes in your configuration profile?

Do you get a getty login prompt on the 2.6.18-x86_64 kernel?

Can you paste the getty lines from your /etc/inittab?

For reference, 2.6.18 kernels want /dev/tty1 for console, whereas the pv_ops kernels (those > 2.6.18) want /dev/hvc0. Xenify is supposed to take care of making this modification for you, even when switching between kernels.

-Chris

@caker:

For reference, 2.6.18 kernels want /dev/tty1 for console, whereas the pv_ops kernels (those > 2.6.18) want /dev/hvc0. Xenify is supposed to take care of making this modification for you, even when switching between kernels. Thank you very much. Didn't know that one.

So no, I didn't have xenification turned on because it wasn't need in 2.6.18_x64.

One more thing - I am missing /proc/xen directory with this new kernel. Is it feature?

It's a completely different implementation of xen guest support.

Just wanted to give the thumbs up from me. Been running it since about the time this thread started and no complaints. Nice to see so much effort being put into keeping us current.

@zunzun:

I don't normally use this kind of arcane technical jargon, but my graphs are definitely poopy-doopy.

I'm now inclined to believe this is munin-specific, and that memory cache is fine.

James

I am apparently having an out-of-memory event. I wouldn't think I'd be susceptible to that, but I don't think I can blame it on the kernel in any case.

However: the OOM killer has blown up. The system is totally unresponsive and the console is repeating this over and over forever:

Mem-info:
Normal per-cpu:
CPU    0: Hot: hi:  186, btch:  31 usd:  12   Cold: hi:   62, btch:  15 usd:  51
Active:84829 inactive:41008 dirty:0 writeback:0 unstable:0
 free:655 slab:3389 mapped:3 pagetables:1201 bounce:0
Normal free:2620kB min:2960kB low:3700kB high:4440kB active:339316kB inactive:164032kB present:548640kB pages_scanned:228230182 all_unreclaimable? yes
lowmem_reserve[]: 0 0
Normal: 1*4kB 1*8kB 17*16kB 11*32kB 1*64kB 1*128kB 1*256kB 1*512kB 1*1024kB 0*2048kB 0*4096kB = 2620kB
Swap cache: add 1158766, delete 1158766, find 2060666/2215923, race 5+107
Free swap  = 0kB
Total swap = 263160kB
Free swap:            0kB
138240 pages of RAM
0 pages of HIGHMEM
3950 reserved pages
573 pages shared
0 pages swap cached
printk: 27627 messages suppressed.
cron invoked oom-killer: gfp_mask=0xa01d2, order=0, oomkilladj=0
1e297cb0:  [<080693f2>] dump_stack+0x22/0x30
1e297cc8:  [<080b3cd9>] out_of_memory+0x109/0x140
1e297cf4:  [<080b51c5>] __alloc_pages+0x355/0x380
1e297d48:  [<080b7688>] __do_page_cache_readahead+0x128/0x1a0
1e297d84:  [<080b77f1>] do_page_cache_readahead+0x51/0x70
1e297da4:  [<080b1615>] filemap_fault+0x1f5/0x310
1e297de8:  [<080bed75>] __do_fault+0x55/0x3d0
1e297e2c:  [<080bf13e>] do_linear_fault+0x4e/0x50
1e297e50:  [<080bf39c>] handle_mm_fault+0xbc/0x2a0
1e297e84:  [<0806a52c>] handle_page_fault+0x13c/0x230
1e297eb8:  [<0806a887>] segv+0x177/0x2f0
1e297f6c:  [<0807e046>] handle_segv+0x56/0x60
1e297f90:  [<0807e856>] userspace+0x216/0x260
1e297fe4:  [<0806b434>] fork_handler+0x74/0x90
1e297ffc:  [<4000b080>] 0x4000b080

My bad. This happened on another of my Linodes which is actually still on UML. Nothing to do with the new Xen kernel which is still working fine on the Linode I have it on.

I've just switched to this new kernel on my Ubuntu 8.04. The only problem was that Apache failed to start on boot. After restarting the daemon everything was fine!

Performance-wise I didn't see much difference. Both RAM and CPU usage are still the same.

Booted into this kernel on my 360 earlier and noticed that it reports my memory incorrectly. If it helps, I'm on dallas111.

free -m -t

and htop reports 349MB of available memory while using the stable kernel, both reports 360MB of memory (as it should).

It was nice though, reporting that I'm only using 35MB/349MB versus 60MB/360MB which was the usual usage on the stable kernel.

Unfortunately, I can't tell if that was just a misreport or actually savings… :P

[root@ ~]# uname -a

Linux x.y.com 2.6.27.4-x86_64-linode3

when running

/sbin/iptables -t mangle -A MANGLE_OUTPUT -p tcp –dport 80 -j TOS --set-tos Maximize-Throughput

the output is

iptables: Unknown error 18446744073709551615

i use this ruleset on other hosts (containing this rule), and haven't had this problem before.

Thanks,

Reply

Please enter an answer
Tips:

You can mention users to notify them: @username

You can use Markdown to format your question. For more examples see the Markdown Cheatsheet.

> I’m a blockquote.

I’m a blockquote.

[I'm a link] (https://www.google.com)

I'm a link

**I am bold** I am bold

*I am italicized* I am italicized

Community Code of Conduct