Folks,
As you''re probably aware we''ve been running a number of
different
initiatives to test the xen -unstable tree prior to release as 3.0.0: We
have the XenRT automated regression test suite which gives Xen a
thorough workout on a range of different machines every night. Among
other tests, XenRT incorporates the excellent xm-test suite to stress
xend. We''re also running various ISV application validation suites over
Xen. There''s also all the testing that''s going on in the
community, and
the daily reports from IBM and Intel. All indications are that we''re
getting very close to release quality.
One aspect that current testing doesn''t provide us with much coverage
on
is the wide range of PC hardware that users wish to run Xen on. We
believe the best way of plugging this gap is with the Xen automated
TestCD, and I''m pleased to announce the first version is now available.
After initial configuration, the CD runs in an unattended fashion and is
specifically designed to check that Xen runs stably on the hardware
platform. It exercises all disk controllers (read-only!) and network
interfaces it finds, along with some simple graphics tests. During the
process it automatically reboots into a number of different kernels and
hypervisor versions, enabling us to compare results across native linux
and x86_32, x86_32 PAE and x86_64 versions of Xen. After completing the
tests it tries to upload a tar file of the results to xensource.com over
http for analysis.
(If you''re running the tests on a farm of machines that don''t
have
external net access there''s an option to store the results file on a
USB
key stick or other removable media and then upload manually via a web
form or via email). The results will get processed by various scripts
and then be available for viewing via a web interface.
[We''re not quite ready to go-live with the web interface, so check back
in a few days -- it''s important we get the data collection underway
though]
It would be very helpful if you could arrange to run the test CD on all
the hardware platforms you''re serious about wanting to run Xen on (i.e.
not that old 486 you have in the cupboard under the stairs :-)
Although the test CD can''t test everything, it should give us a decent
idea of what hardware Xen runs successfully on today, and draw attention
to particular drivers or chipsets we have problems with.
You can download the CD from here:
http://www.xensource.com/community/xen30testing/downloadcd.html
NB: To enable the network tests to run you''ll need to run a Network
Test
Server". Either download the separate test server package from above web
page, or burn a second CD, boot it in the server and select the ''test
server'' option from the grub menu. The machine under test will use
multicast DNS to rendezvous with the server and bounce traffic off it.
If you''re behind a firewall that won''t let http through
you''ll need to
upload the results tar file manually via the following web page:
http://www.xensource.com/community/xen30testing/results_form.php
Many thanks for your support in ensuring a high quality Xen 3.0.0
release!
Best,
Ian
PS: Here''s the release notes for the CD, also available on the web
site.
=========================================================================
Here''s a quick description of how the Xen 3.0 Test CD works:
The CD will first boot into a setup environment that lets you choose a
partition on detected hard disks on which it will store test results and
some config files that it will use to proceed through the various test
runs. It will create a directory called /xentestingcd on the chosen
partition.
Depending on the type of machine, the test will automatically reboot
several times to run the next phase of the tests. (A grub menu is
displayed, but you don''t need to interact.)
On completion of the tests, the CD will attempt to submit the results to
our website via http.
Note that the results will be also be left on the hard disk in the
/xentestingcd directory, in case the attempt to automatically upload the
results fails-for example, when there is no connection to the internet
available. If you can, it would be great if you could copy that file
(named xenresults.tar.bz2.abcde, where the extension is a
randomly-generated string of characters) somewhere that you can upload
it to us via the form [link]here[/link].
For a more thorough discussion of the Test CD''s operation and a list of
known issues, see the [link]Release Notes[/link].
==============================================
The Release Notes are as follows:
==============================================
Xen 3.0 Test CD Release Notes
The Xen test CD is designed to test the hardware of a machine for
compatibility issues with Xen, and report any problems found to the Xen
development team so that any problems that may exist can be fixed based
up on a large pool of test data.
Requirements
In order to successfully run the CD, you must have a system that has
either a disk with an ext2 or ext3 filesystem that is not managed by
LVM, or a USB key that is visible to the computer BIOS. This is because
the CD uses GRUB to boot Linux and Xen, and uses state saved at a
location you specify on the initial boot in order to determine which
kernel or Xen build should be booted.
Instructions For Use
Network tests
The test CD has a number of network tests that require a network test
server to be present on the network, and on the same subnet as the
machines that are being tested.
In order to run a network test server, one should boot the CD from a
machine that is not being tested, and choose the Network Test Server
option from the GRUB boot menu (choosing either vga or serial version as
required).
Alternatively, one can download and run a stand-alone network test
server on from a machine that is not being tested and is on the same
subnet. See the Text CD Download page.
Booting the Test CD
When you first boot the CD, you will be presented with a GRUB menu. At
this point, you should determine whether you would like to use a serial
or VGA console, and make a selection of either "Setup XTCD (vga)" or
"Setup XTCD (serial)." The default selection is "Setup XTCD
(vga)."
The CD will now boot into a native Linux kernel, and will ask for a
location where tests results can be saved, and a second location (which
is often the same as the first, unless the first is a USB storage
device) where state may be saved.
NOTE-It is important that the state location be visible to the computer
BIOS in order for booting of the CD to operate correctly.
The CD can be left to run unattended once these selections have been
made, until the "Submit Results" stage, where you will be prompted for
further information. When you see the GRUB menu, just leaving
''Default''
as the selection should suffice. Choosing "Default" will cause a
second
GRUB menu to be loaded, where the default selection will automatically
change as the tests progress. This menu is stored on the state drive,
hence the requirement for this drive to be visible to BIOS.
If something goes wrong
The CD uses two mechanisms for ensuring that useful results can be
gained even if a crash occurs partway though testing:
* If you manually reboot the CD, choosing the Default option at the
initial GRUB prompt will cause the CD to boot into the next kernel to be
tested rather than looping to the one that was last used.
* Some scripts have timeouts on them, so if they hang then other
tests may still run.
Running the Submit Results stage is important because, even if you chose
not to submit results to XenSource for analysis (you have the option not
to when running this stage), it will clean up any state left behind and
leave with a .tar.bz2 file containing any logs that were generated
during execution of the test CD.
Known Issues
* Currently the network tests that are performed in guest domains
only work if DHCP can be used to configure networking within the domain.
*
* The test CD might fail on some machines with SATA interfaces.
* The NTFS and UFS file systems are not supported for writing data
results.
Version Information
The CD has a version label as follows:
xen-3.0-testCD-[pre]<major>.<minor>[-(rev)]
The version of the CD in use can be found in the REVISION file in the
root of the CD''s filesystem.
The version of the CD used to create a results tarball is in a REVISION
file located inside the tarball - this is copied directly from the CD
when the results tarball is created.
=============================================
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
While the CD seems to boot and work fine in my thinkpad, it just hangs very
early at the
Loading stage2
Press any key
stage when I try to boot it in my dual opteron Tyan k8w workstation.
Do I perhaps fall into the
* The test CD might fail on some machines with SATA interfaces.
category?
Andrew Walrond
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
> While the CD seems to boot and work fine in my thinkpad, it > just hangs very early at the > > Loading stage2 > Press any key > > stage when I try to boot it in my dual opteron Tyan k8w workstation. > > Do I perhaps fall into the > > * The test CD might fail on some machines with SATA interfaces. > > category?It sounds like your machine is failing at the grub stage, which we haven''t encountered before except on very old machines with dodgy bios''es. syslinux definitely seems to work on more machines than grub, but on a modern motherboard like yours I''m surprised there''s a problem. I don''t suppose there''s a bios upgrade available for your machine? Ian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Ian, Can you comment if the bugs I''ve reported in bugzilla #''s 402, 405, and 406 have been fixed/assigned? I tried to change a few of our (unixshell.com) users over to a 3.0 server and it turned out poorly. We have serial console (with logging) setup for our servers so when a system panics or sends an oops so I am more than capable of capturing it submitting bugs for any troubles, but I received those different 3 panics and the system couldn''t stay up for more than a few hours (I had even tried turning off PAE and SMP support). This was all using snapshots from around 11/13. I know there have been many, many commits since then. Thanks, Matt Ian Pratt wrote:> Folks, > > As you''re probably aware we''ve been running a number of different > initiatives to test the xen -unstable tree prior to release as 3.0.0: We > have the XenRT automated regression test suite which gives Xen a > thorough workout on a range of different machines every night. Among > other tests, XenRT incorporates the excellent xm-test suite to stress > xend. We''re also running various ISV application validation suites over > Xen. There''s also all the testing that''s going on in the community, and > the daily reports from IBM and Intel. All indications are that we''re > getting very close to release quality. >_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser
2005-Nov-25 17:39 UTC
Re: [Xen-devel] Re: [Xen-users] xen 3.0 Test CD v1 released
On 25 Nov 2005, at 17:33, Matt Ayres wrote:> Can you comment if the bugs I''ve reported in bugzilla #''s 402, 405, > and 406 have been fixed/assigned? I tried to change a few of our > (unixshell.com) users over to a 3.0 server and it turned out poorly. > We have serial console (with logging) setup for our servers so when a > system panics or sends an oops so I am more than capable of capturing > it submitting bugs for any troubles, but I received those different 3 > panics and the system couldn''t stay up for more than a few hours (I > had even tried turning off PAE and SMP support). This was all using > snapshots from around 11/13. I know there have been many, many > commits since then.402 and 405 are the same bug. Can you really repro it with a non-PAE kernel? It looks related to using swiotlb to access memory >4GB. If you can repro 406 with a debug build of Xen (debug=y make) then the backtrace will make much more sense. Cheers, Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On Friday 25 November 2005 00:34, Ian Pratt wrote:> > It sounds like your machine is failing at the grub stage, which we > haven''t encountered before except on very old machines with dodgy > bios''es. syslinux definitely seems to work on more machines than grub, > but on a modern motherboard like yours I''m surprised there''s a problem. > I don''t suppose there''s a bios upgrade available for your machine?I boot the machine every day into linux with grub, so this is a bit strange. I can also boot my own grub-bootable CDs just fine. I''ll see if I can investigate further... Andrew Walrond _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On 11/25/05, Andrew Walrond <andrew@walrond.org> wrote:> I boot the machine every day into linux with grub, so this is a bit strange. I > can also boot my own grub-bootable CDs just fine. > I''ll see if I can investigate further...I am having what might be a similar problem on my Xen machine. Grub loads for me without hanging, but I just end up at the grub> prompt, no menu. I can''t load the menu.lst manually because it''s not an ascii text file. I just ran an md5sum on my iso, and it doesn''t match up with what''s on the xensource page at all. I downloaded that iso from the Pennsylvania mirror. I''m going to try the Texas one now and see if that works.. You might want to see if your iso is corrupt also. Dave -- David Muench - davemuench@gmail.com Jabber ID: dave@jabber.wasteland.org _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On 11/25/05, David Muench <davemuench@gmail.com> wrote:> I just ran an md5sum on my iso, and it doesn''t match up with what''s on > the xensource page at all. I downloaded that iso from the Pennsylvania > mirror. I''m going to try the Texas one now and see if that works.. You > might want to see if your iso is corrupt also.The iso downloaded from Texas has the appropriate md5sum. I''m burning it now. The Pennsylvania one should be double checked by the xensource guys. Dave -- David Muench - davemuench@gmail.com Jabber ID: dave@jabber.wasteland.org _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On 11/25/05, David Muench <davemuench@gmail.com> wrote:> The iso downloaded from Texas has the appropriate md5sum. I''m burning > it now. The Pennsylvania one should be double checked by the xensource > guys.New problems - the cd booted fine this time, and went through the initial battery of tests in normal Linux without any problems. When it rebooted to go into Xen 3.0 32 bit, the first grub screen (where it picked Default) worked fine, but the second one to actually boot Xen didn''t. Here''s the error (retyped): Booting ''Xen 3.0 (x86_32) (vga)'' savedefault 1 128:393215 /xentestingcd/default root (cd) Error 25: Disk read error Press any key to continue... It''s the root (cd) causing the disk read error, I got the same thing when trying to run that command at the grub command line. My cdrom is a standard PATA hooked up as secondary master. Dave -- David Muench - davemuench@gmail.com Jabber ID: dave@jabber.wasteland.org _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> On 11/25/05, David Muench <davemuench@gmail.com> wrote: > > I just ran an md5sum on my iso, and it doesn''t match up > with what''s on > > the xensource page at all. I downloaded that iso from the > Pennsylvania > > mirror. I''m going to try the Texas one now and see if that > works.. You > > might want to see if your iso is corrupt also. > > The iso downloaded from Texas has the appropriate md5sum. I''m > burning it now. The Pennsylvania one should be double checked > by the xensource guys.I''ve downloaded the iso from the Penn mirror and checked the md5sum and it looks OK. Are you sure it wasn''t truncated when you pulled it? The size is 530554880 bytes. Thank, Ian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Nov 25, 2005, at 3:42 PM, David Muench wrote:> On 11/25/05, David Muench <davemuench@gmail.com> wrote: > >> The iso downloaded from Texas has the appropriate md5sum. I''m burning >> it now. The Pennsylvania one should be double checked by the >> xensource >> guys. > > New problems - the cd booted fine this time, and went through the > initial battery of tests in normal Linux without any problems. When it > rebooted to go into Xen 3.0 32 bit, the first grub screen (where it > picked Default) worked fine, but the second one to actually boot Xen > didn''t. Here''s the error (retyped): > > Booting ''Xen 3.0 (x86_32) (vga)'' > > savedefault 1 > 128:393215 /xentestingcd/default > root (cd) > > Error 25: Disk read error > > Press any key to continue... > > It''s the root (cd) causing the disk read error, I got the same thing > when trying to run that command at the grub command line. My cdrom is > a standard PATA hooked up as secondary master.Dave, I got the same reaction when I first tried it. I was using one of the later partitions for the XenTest data. When I changed that to the first partition (a /boot area that has extra space) everything worked as expected. So I''m wondering if the error is really coming from an attempt to read the HD for test status. Richard _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Friday 25 November 2005 19:43, David Muench wrote:> > I just ran an md5sum on my iso, and it doesn''t match up with what''s on > the xensource page at all. I downloaded that iso from the Pennsylvania > mirror. I''m going to try the Texas one now and see if that works.. You > might want to see if your iso is corrupt also. >This is not the same problem; the CD boots just fine on other machines. Andrew _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Friday 25 November 2005 18:24, Andrew Walrond wrote:> On Friday 25 November 2005 00:34, Ian Pratt wrote: > > It sounds like your machine is failing at the grub stage, which we > > haven''t encountered before except on very old machines with dodgy > > bios''es. syslinux definitely seems to work on more machines than grub, > > but on a modern motherboard like yours I''m surprised there''s a problem. > > I don''t suppose there''s a bios upgrade available for your machine? > > I boot the machine every day into linux with grub, so this is a bit > strange. I can also boot my own grub-bootable CDs just fine. > I''ll see if I can investigate further... >A bios upgrade did indeed solve the problem. The tests ran smoothly, and the results were (apparently; no email confirmation yet) uploaded to your server. I hope the results are useful. I have two other dual opteron servers I will be testing shortly, with different m/b + hardware. Andrew Walrond _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> Dave, > > I got the same reaction when I first tried it. I was using > one of the later partitions for the XenTest data. > When I changed that to the first partition (a /boot area that > has extra space) everything worked as expected. > > So I''m wondering if the error is really coming from an > attempt to read the HD for test status. > > RichardWe now understand this issue with the CD, and it turns out to be a bug in grub (bug #14407). Grub''s partition table handling doesn''t reset itself properly when going from an "extended" DOS partition on a HDD to accessing a CD, and ends up trying to read from the CD at the same offset as the logical partition table was on the drive. :( Workarounds: - Don''t use an extended partition to save the state directory. - Edit the boot commands to issue "root (hd0,0)" just before "root (cd)", which should unstick the partition table pointer. You''d have to do this on every boot, though. :( Thanks, Ian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Thursday 24 November 2005 14:18, Ian Pratt wrote:> > It would be very helpful if you could arrange to run the test CD on all > the hardware platforms you''re serious about wanting to run Xen on (i.e. > not that old 486 you have in the cupboard under the stairs :-) >Should I have received an email confirming receipt of the test results? I assumed I would since I was asked for my email address. So now I''m wondering whether the results actually got sent, or whether I should do a manual upload. Guidance appreciated, before I test the next batch of machines. Thanks Andrew Walrond _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> On Thursday 24 November 2005 14:18, Ian Pratt wrote: > > > > It would be very helpful if you could arrange to run the test CD on > > all the hardware platforms you''re serious about wanting to > run Xen on (i.e. > > not that old 486 you have in the cupboard under the stairs :-) > > > > Should I have received an email confirming receipt of the > test results? I assumed I would since I was asked for my > email address. So now I''m wondering whether the results > actually got sent, or whether I should do a manual upload. > > Guidance appreciated, before I test the next batch of machines.We hadn''t actually thought of sending a ''results received OK'' email, but it''s a good idea. (We were collecting the email address mainly so that we could contact people who had machines exhibiting bugs and ask them to run follow up tests) I think what we''ll do is modify the script that processes the results to generate a summary and then email that back to the submitter. Thanks for the feedback! Ian _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Matt Ayres
2005-Nov-27 20:35 UTC
Re: [Xen-devel] Re: [Xen-users] xen 3.0 Test CD v1 released
Keir Fraser wrote:> > On 25 Nov 2005, at 17:33, Matt Ayres wrote: > >> Can you comment if the bugs I''ve reported in bugzilla #''s 402, 405, >> and 406 have been fixed/assigned? I tried to change a few of our >> (unixshell.com) users over to a 3.0 server and it turned out poorly. >> We have serial console (with logging) setup for our servers so when a >> system panics or sends an oops so I am more than capable of capturing >> it submitting bugs for any troubles, but I received those different 3 >> panics and the system couldn''t stay up for more than a few hours (I >> had even tried turning off PAE and SMP support). This was all using >> snapshots from around 11/13. I know there have been many, many >> commits since then. > > 402 and 405 are the same bug. Can you really repro it with a non-PAE > kernel? It looks related to using swiotlb to access memory >4GB. >I should have documented this better, I believe due to the panics in 402/405 I switched to non-PAE, non-SMP dom0 and domU kernels and then produced 406.> If you can repro 406 with a debug build of Xen (debug=y make) then the > backtrace will make much more sense. >I can try to do this. Of course I can''t guarantee to get the panic, but if I get do see anything I''ll make sure the kernels are built with debugging enabled. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> Should I have received an email confirming receipt of the > test results? I assumed I would since I was asked for my > email address. So now I''m wondering whether the results > actually got sent, or whether I should do a manual upload. > > Guidance appreciated, before I test the next batch of machines.Andrew, Your results were indeed uploaded; there is feedback on the screen at the point it''s uploaded that it succeeded (or not). But an acknowledgement email is indeed a good idea, as Ian already noted, and I''ll implement that within the next day. -Alan _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Fraser Campbell
2005-Nov-28 16:54 UTC
[Xen-devel] Re: [Xen-users] xen 3.0 Test CD v1 released
Ian Pratt wrote:> After initial configuration, the CD runs in an unattended fashion and is > specifically designed to check that Xen runs stably on the hardware > platform.Great stuff, ran the tests on an AMD64 system at home and everything appeared to work though I didn''t look at the results file (total runtime about 50 minutes). Will give it a go on some HP and IBM servers that we have in the office. Are you guys still maintaining the Cambridge site? It would probably be a good idea to update http://www.cl.cam.ac.uk/Research/SRG/netos/xen/downloads.html to mention the availability of this CD so that more people will use it. Thanks! Fraser _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 11/26/05, Ian Pratt <m+Ian.Pratt@cl.cam.ac.uk> wrote:> - Edit the boot commands to issue "root (hd0,0)" just before "root > (cd)", > which should unstick the partition table pointer. You''d have to do > this on every boot, though. :(This method worked fine today, and I was able to successfully run the test CD in 32 and 32pae modes on my Xen box. Just uploaded the results. Xen 3.0 looks like it''s going to be great. Dave -- David Muench - davemuench@gmail.com Jabber ID: dave@jabber.wasteland.org _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Chris Bainbridge
2005-Nov-30 11:51 UTC
Re: [Xen-devel] Re: [Xen-users] xen 3.0 Test CD v1 released
Hi, There is a bug running the CD as a network server: /sbin/testserver.sh : line 16: dhcpcd not found Also, it failed to boot on SATA but from the release notes it seems that''s a known problem. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
I found a small bug in the Test CD. Even if you configured a
proxy server, it wasn''t being used in the submission script.
It''s a one-line patch to xtcdsubmit.pl.
Now to respin the CD. Does anyone have quick instructions?
ja.
--
Joe Ammond joe.ammond@edwardjones.com
plug into her electric cool
where things bend and break and shake to the rules
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel