Osstest: stop testing SEDF, start testing RTDS
PATCH 1/1 Osstest: stop testing SEDF, start testing RTDS
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
It therefore makes sense to stop smoke testing the former in favour of the latter. Note that the -sedf-pin jobs where only added in order to try to debug a long standing issue with SEDF; it is not necessary to have anything like that for RTDS.
For now, as RTDS is still marked as experimental, test failures are allowed, as it was for SEDF.
[1] http://lists.xen.org/archives/html/xen-devel/2015-05/msg02874.html
--- allow.all | 2 +- make-flight | 31 +++++++++---------------------- 2 files changed, 10 insertions(+), 23 deletions(-)
allow.all +1 -1
@@ -1,4 +1,4 @@ -test-@@-sedf@@ +test-@@-rtds@@ build-@@ logs-capture@@ test-@@-pcipt@@ test-@@-qemuu-@@ guest-localmigrate
make-flight +9 -22
@@ -262,30 +262,16 @@ do_hvm_rhel6_tests () { done } -do_sedf_tests () { +do_nondef_sched_tests () { if [ $xenarch != $dom0arch ]; then return fi - for pin in '' -pin; do - job_create_test test-$xenarch$kern-$dom0arch-xl-sedf$pin \ - test-debian xl $xenarch $dom0arch \ - guests_vcpus=4 \ - xen_boot_append="sched=sedf loglvl=all ${pin:+dom0_vcpus_pin}" \ - linux_boot_append='loglevel=9 debug' \ - $debian_runvars all_hostflags=$most_hostflags - done -} - -do_credit2_tests () { - if [ $xenarch != $dom0arch ]; then - return - fi - - job_create_test test-$xenarch$kern-$dom0arch-xl-credit2 \ - test-debian xl $xenarch $dom0arch \ - guests_vcpus=4 xen_boot_append='sched=credit2' \ - $debian_runvars all_hostflags=$most_hostflags + sched=$1 + job_create_test test-$xenarch$kern-$dom0arch-xl-$sched \ + test-debian xl $xenarch $dom0arch \ + guests_vcpus=4 xen_boot_append="sched=$sched" \ + $debian_runvars all_hostflags=$most_hostflags } do_multivcpu_tests () {
@@ -350,8 +336,9 @@ test_matrix_do_one () { do_pv_debian_tests do_multivcpu_tests - do_sedf_tests - do_credit2_tests + for s in credit2 rtds; do + do_nondef_sched_tests $s + done # No further arm tests at the moment if [ $dom0arch = armhf ]; then
Signature · 4 lines
<<This happens because I choose it to happen!>> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
Quoted 9 lines · Dario Faggioli
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
It therefore makes sense to stop smoke testing the former in favour of the latter. Note that the -sedf-pin jobs where only added in order to try to debug a long standing issue with SEDF; it is not necessary to have anything like that for RTDS.
Does the RTDS scheduler work out of the box? I had a feeling that it required per domain config settings to be worked out?
Quoted 76 lines
For now, as RTDS is still marked as experimental, test failures are allowed, as it was for SEDF.
[1] http://lists.xen.org/archives/html/xen-devel/2015-05/msg02874.html
Signed-off-by Dario Faggioli <[email protected]>--- allow.all | 2 +- make-flight | 31 +++++++++---------------------- 2 files changed, 10 insertions(+), 23 deletions(-)allow.all +1
-1@@ -1,4 +1,4 @@ -test-@@-sedf@@ +test-@@-rtds@@ build-@@ logs-capture@@ test-@@-pcipt@@ test-@@-qemuu-@@ guest-localmigratemake-flight +9
-22@@ -262,30 +262,16 @@ do_hvm_rhel6_tests () { done } -do_sedf_tests () { +do_nondef_sched_tests () { if [ $xenarch != $dom0arch ]; then return fi - for pin in '' -pin; do - job_create_test test-$xenarch$kern-$dom0arch-xl-sedf$pin \ - test-debian xl $xenarch $dom0arch \ - guests_vcpus=4 \ - xen_boot_append="sched=sedf loglvl=all ${pin:+dom0_vcpus_pin}" \ - linux_boot_append='loglevel=9 debug' \ - $debian_runvars all_hostflags=$most_hostflags - done -} - -do_credit2_tests () { - if [ $xenarch != $dom0arch ]; then - return - fi - - job_create_test test-$xenarch$kern-$dom0arch-xl-credit2 \ - test-debian xl $xenarch $dom0arch \ - guests_vcpus=4 xen_boot_append='sched=credit2' \ - $debian_runvars all_hostflags=$most_hostflags + sched=$1 + job_create_test test-$xenarch$kern-$dom0arch-xl-$sched \ + test-debian xl $xenarch $dom0arch \ + guests_vcpus=4 xen_boot_append="sched=$sched" \ + $debian_runvars all_hostflags=$most_hostflags } do_multivcpu_tests () {@@ -350,8 +336,9 @@ test_matrix_do_one () { do_pv_debian_tests do_multivcpu_tests - do_sedf_tests - do_credit2_tests + for s in credit2 rtds; do + do_nondef_sched_tests $s + done # No further arm tests at the moment if [ $dom0arch = armhf ]; then
Quoted 5 lines · Ian Campbell
Quoted 4 lines · Dario Faggioli
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
Quoted 9 lines
Quoted 5 lines
It therefore makes sense to stop smoke testing the former in favour of the latter. Note that the -sedf-pin jobs where only added in order to try to debug a long standing issue with SEDF; it is not necessary to have anything like that for RTDS.
Does the RTDS scheduler work out of the box? I had a feeling that it required per domain config settings to be worked out?
It does, your feeling is about ARINC653 :-)
Regards,
Dario
Signature · 4 lines
<<This happens because I choose it to happen!>> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
Quoted 12 lines · Dario Faggioli
Quoted 11 lines · Ian Campbell
Quoted 9 lines · Dario Faggioli
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
It therefore makes sense to stop smoke testing the former in favour of the latter. Note that the -sedf-pin jobs where only added in order to try to debug a long standing issue with SEDF; it is not necessary to have anything like that for RTDS.
Quoted 4 lines
Quoted 3 lines
Does the RTDS scheduler work out of the box? I had a feeling that it required per domain config settings to be worked out?
It does, your feeling is about ARINC653 :-)
Great, thanks!
Ian.
[ Adding to Cc the people that I though I added when sending the patch,
but that apparently I haven't...sorry :-( ]
Quoted 10 lines · Dario Faggioli
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
It therefore makes sense to stop smoke testing the former in favour of the latter. Note that the -sedf-pin jobs where only added in order to try to debug a long standing issue with SEDF; it is not necessary to have anything like that for RTDS.
Quoted 5 lines
For now, as RTDS is still marked as experimental, test failures are allowed, as it was for SEDF.
[1] http://lists.xen.org/archives/html/xen-devel/2015-05/msg02874.html
Oh, and here's the diff of the generated runvars that IanJ usually asks
about: :-)
dariof@drall:~/osstest.git$ diff -Nru runvar.prepatch runvar.patched --- runvar.prepatch 2015-05-21 17:13:51.328281000 +0100 +++ runvar.patched 2015-05-21 17:15:51.112202000 +0100 @@ -370,30 +370,17 @@ test-amd64-amd64-xl-qemuu-winxpsp3 win_acpi_shutdown true test-amd64-amd64-xl-qemuu-winxpsp3 win_image winxpsp3.iso test-amd64-amd64-xl-qemuu-winxpsp3 xenbuildjob build-amd64 -test-amd64-amd64-xl-sedf all_hostflags arch-amd64,arch-xen-amd64,suite-wheezy,purpose-test -test-amd64-amd64-xl-sedf arch amd64 -test-amd64-amd64-xl-sedf buildjob build-amd64 -test-amd64-amd64-xl-sedf debian_arch amd64 -test-amd64-amd64-xl-sedf debian_kernkind pvops -test-amd64-amd64-xl-sedf guests_vcpus 4 -test-amd64-amd64-xl-sedf kernbuildjob build-amd64-pvops -test-amd64-amd64-xl-sedf kernkind pvops -test-amd64-amd64-xl-sedf linux_boot_append loglevel=9 debug -test-amd64-amd64-xl-sedf-pin all_hostflags arch-amd64,arch-xen-amd64,suite-wheezy,purpose-test -test-amd64-amd64-xl-sedf-pin arch amd64 -test-amd64-amd64-xl-sedf-pin buildjob build-amd64 -test-amd64-amd64-xl-sedf-pin debian_arch amd64 -test-amd64-amd64-xl-sedf-pin debian_kernkind pvops -test-amd64-amd64-xl-sedf-pin guests_vcpus 4 -test-amd64-amd64-xl-sedf-pin kernbuildjob build-amd64-pvops -test-amd64-amd64-xl-sedf-pin kernkind pvops -test-amd64-amd64-xl-sedf-pin linux_boot_append loglevel=9 debug -test-amd64-amd64-xl-sedf-pin toolstack xl -test-amd64-amd64-xl-sedf-pin xen_boot_append sched=sedf loglvl=all dom0_vcpus_pin -test-amd64-amd64-xl-sedf-pin xenbuildjob build-amd64 -test-amd64-amd64-xl-sedf toolstack xl -test-amd64-amd64-xl-sedf xen_boot_append sched=sedf loglvl=all -test-amd64-amd64-xl-sedf xenbuildjob build-amd64 +test-amd64-amd64-xl-rtds all_hostflags arch-amd64,arch-xen-amd64,suite-wheezy,purpose-test +test-amd64-amd64-xl-rtds arch amd64 +test-amd64-amd64-xl-rtds buildjob build-amd64 +test-amd64-amd64-xl-rtds debian_arch amd64 +test-amd64-amd64-xl-rtds debian_kernkind pvops +test-amd64-amd64-xl-rtds guests_vcpus 4 +test-amd64-amd64-xl-rtds kernbuildjob build-amd64-pvops +test-amd64-amd64-xl-rtds kernkind pvops +test-amd64-amd64-xl-rtds toolstack xl +test-amd64-amd64-xl-rtds xen_boot_append sched=rtds +test-amd64-amd64-xl-rtds xenbuildjob build-amd64 test-amd64-amd64-xl toolstack xl test-amd64-amd64-xl xenbuildjob build-amd64 test-amd64-amd64-xl-xsm all_hostflags arch-amd64,arch-xen-amd64,suite-wheezy,purpose-test @@ -691,30 +678,17 @@ test-armhf-armhf-xl-multivcpu kernkind pvops test-armhf-armhf-xl-multivcpu toolstack xl test-armhf-armhf-xl-multivcpu xenbuildjob build-armhf -test-armhf-armhf-xl-sedf all_hostflags arch-armhf,arch-xen-armhf,suite-wheezy,purpose-test -test-armhf-armhf-xl-sedf arch armhf -test-armhf-armhf-xl-sedf buildjob build-armhf -test-armhf-armhf-xl-sedf debian_arch armhf -test-armhf-armhf-xl-sedf debian_kernkind pvops -test-armhf-armhf-xl-sedf guests_vcpus 4 -test-armhf-armhf-xl-sedf kernbuildjob build-armhf-pvops -test-armhf-armhf-xl-sedf kernkind pvops -test-armhf-armhf-xl-sedf linux_boot_append loglevel=9 debug -test-armhf-armhf-xl-sedf-pin all_hostflags arch-armhf,arch-xen-armhf,suite-wheezy,purpose-test -test-armhf-armhf-xl-sedf-pin arch armhf -test-armhf-armhf-xl-sedf-pin buildjob build-armhf -test-armhf-armhf-xl-sedf-pin debian_arch armhf -test-armhf-armhf-xl-sedf-pin debian_kernkind pvops -test-armhf-armhf-xl-sedf-pin guests_vcpus 4 -test-armhf-armhf-xl-sedf-pin kernbuildjob build-armhf-pvops -test-armhf-armhf-xl-sedf-pin kernkind pvops -test-armhf-armhf-xl-sedf-pin linux_boot_append loglevel=9 debug -test-armhf-armhf-xl-sedf-pin toolstack xl -test-armhf-armhf-xl-sedf-pin xen_boot_append sched=sedf loglvl=all dom0_vcpus_pin -test-armhf-armhf-xl-sedf-pin xenbuildjob build-armhf -test-armhf-armhf-xl-sedf toolstack xl -test-armhf-armhf-xl-sedf xen_boot_append sched=sedf loglvl=all -test-armhf-armhf-xl-sedf xenbuildjob build-armhf +test-armhf-armhf-xl-rtds all_hostflags arch-armhf,arch-xen-armhf,suite-wheezy,purpose-test +test-armhf-armhf-xl-rtds arch armhf +test-armhf-armhf-xl-rtds buildjob build-armhf +test-armhf-armhf-xl-rtds debian_arch armhf +test-armhf-armhf-xl-rtds debian_kernkind pvops +test-armhf-armhf-xl-rtds guests_vcpus 4 +test-armhf-armhf-xl-rtds kernbuildjob build-armhf-pvops +test-armhf-armhf-xl-rtds kernkind pvops +test-armhf-armhf-xl-rtds toolstack xl +test-armhf-armhf-xl-rtds xen_boot_append sched=rtds +test-armhf-armhf-xl-rtds xenbuildjob build-armhf test-armhf-armhf-xl toolstack xl test-armhf-armhf-xl xenbuildjob build-armhf test-armhf-armhf-xl-xsm all_hostflags arch-armhf,arch-xen-armhf,suite-wheezy,purpose-test
Signature · 4 lines
<<This happens because I choose it to happen!>> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
Hi Dario,
2015-05-22 3:19 GMT-07:00 Dario Faggioli <[email protected]>:
Quoted 3 lines
[ Adding to Cc the people that I though I added when sending the patch,
but that apparently I haven't...sorry :-( ]
Thanks for cc.ing us. :-)
Quoted 12 lines · Dario Faggioli
Quoted 10 lines
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
It therefore makes sense to stop smoke testing the former in favour of the latter. Note that the -sedf-pin jobs where only added in order to try to debug a long standing issue with SEDF; it is not necessary to have anything like that for RTDS.
Quoted 7 lines
Quoted 5 lines
For now, as RTDS is still marked as experimental, test failures are allowed, as it was for SEDF.
[1] http://lists.xen.org/archives/html/xen-devel/2015-05/msg02874.html
Oh, and here's the diff of the generated runvars that IanJ usually asks
about: :-)
I'm not so familiar with osstest. (Maybe I should have a look at it so that we can know how to use it for some auto tests.)
I just have a quick (probably dummy) question:
Do we have any document that describes the coverage of the test?
The reason I asked this is:
1) The document could make it easier for users to know which functionalities of the feature (i.e., RTDS scheduler here) have been tested.
2) I didn't see the toolstack functionality of RTDS scheduler is called in the test. So I'm wondering about the first question. :-) (probably I missed something?)
Thanks,
Meng
-----------
Meng Xu
PhD Student in Computer and Information Science
University of Pennsylvania
http://www.cis.upenn.edu/~mengxu/
Quoted 2 lines · Meng Xu
Hi Dario,
Hi,
Quoted 9 lines
2015-05-22 3:19 GMT-07:00 Dario Faggioli <[email protected]>:
Quoted 6 lines · Dario Faggioli
Quoted 4 lines
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
Quoted 5 lines
I'm not so familiar with osstest. (Maybe I should have a look at it so that we can know how to use it for some auto tests.)
I just have a quick (probably dummy) question:
Do we have any document that describes the coverage of the test?
If, by document, you mean the documentation one would need to used OSSTest, do development on it, and maybe setup and run an instance of, there is some docs --certainly much more than even a few weeks before now:
README files in the sources}
https://blog.xenproject.org/2013/02/02/xen-automatic-test-system-osstest/
http://www.xenproject.org/component/allvideoshare/video/xpds13-osstest.html
https://blog.xenproject.org/2013/09/30/osstest-standalone-mode-step-by-step/
Setting up your own instance would be rather tricky, though, at least untile you get used enough to it.
Quoted 6 lines
The reason I asked this
is:
1) The document could make it easier for users to know which functionalities of the feature (i.e., RTDS scheduler here) have been tested.
I don't think that's how OSSTest works or how it should be used. OSSTest is meant for us developers, in order to be sure that we do not cause regressions. That's why it gates commits.
Users, if we say (e.g., in some release notes), that a feature is 'ready'/\supported\/whatever should be able to use it fully, and complain if it doesn't work, regardless of what tests us developers or OSSTest have run!
Quoted 4 lines
2) I didn't see the toolstack functionality of RTDS scheduler is called in the test. So I'm wondering about the first question. :-) (probably I missed something?)
The test does what it's being done already for other schedulers: sets RTDS as the default scheduler, boots Xen with it, then it creates a domain, and it starts and stop it a few times.
If you want to have a look, check make-flight, sg-run-job, ts-guest-start and ts-guest-stop in OSSTest's codebase.
We do not do any testing of changing scheduling parameters of a guest, if that was what you were referring to, for any of the schedulers. We of course can introduce that (and do it for everyone), but that would be another patch series.
Regards,
Dario
Signature · 4 lines
<<This happens because I choose it to happen!>> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel
Quoted 10 lines · Dario Faggioli
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
It therefore makes sense to stop smoke testing the former in favour of the latter. Note that the -sedf-pin jobs where only added in order to try to debug a long standing issue with SEDF; it is not necessary to have anything like that for RTDS.
Quoted 6 lines
For now, as RTDS is still marked as experimental, test failures are allowed, as it was for SEDF.
[1] http://lists.xen.org/archives/html/xen-devel/2015-05/msg02874.html
Signed-off-by Dario Faggioli <[email protected]>
Dario Faggioli writes ("[PATCH] Osstest: stop testing SEDF, start testing RTDS"):
Quoted 3 lines
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
The RTDS scheduler was introduced in 4.4, AFAICT.
osstest needs to test older versions of Xen too. So I think you need to gate this on the xen version.
Ian.
Quoted 10 lines · Ian Jackson
Dario Faggioli writes ("[PATCH] Osstest: stop testing SEDF, start testing RTDS"):
Quoted 3 lines
the SEDF scheduler is about to be deprecated and go away (see [1]). OTOH, the RTDS scheduler is here to stay.
The RTDS scheduler was introduced in 4.4, AFAICT.
osstest needs to test older versions of Xen too. So I think you need to gate this on the xen version.
Good point, and sorry for missing this! I will respin ASAP.
Dario
Signature · 4 lines
<<This happens because I choose it to happen!>> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
_______________________________________________
Xen-devel mailing list
[email protected]
http://lists.xen.org/xen-devel