Re: [PATCH v3 4/4] tests/qtest/migration: add postcopy test with multifd

[Peter Xu]

On Fri, Jan 24, 2025 at 06:15:20PM +0530, Prasad Pandit wrote:
> Hello Peter,
> 
> On Thu, 23 Jan 2025 at 16:39, Prasad Pandit <ppandit@redhat.com> wrote:
> > On Wed, 22 Jan 2025 at 21:40, Peter Xu <peterx@redhat.com> wrote:
> > > I believe many of the tests can be avoided, but still below is a list of
> > > minimum tests that I think might still be good to add:
> > >         # /x86_64/migration/postcopy/plain
> > >         # /x86_64/migration/postcopy/recovery/tls/psk
> > >         # /x86_64/migration/postcopy/preempt/plain
> > >         # /x86_64/migration/postcopy/preempt/recovery/tls/psk
> ---
> $ ../qtest/migration-test --tap -k -r
> '/x86_64/migration/multifd+postcopy' | grep -i 'slow test'
> # slow test /x86_64/migration/multifd+postcopy/plain executed in 1.28 secs
> # slow test /x86_64/migration/multifd+postcopy/recovery/tls/psk
> executed in 2.43 secs
> # slow test /x86_64/migration/multifd+postcopy/preempt/plain executed
> in 1.52 secs
> # slow test /x86_64/migration/multifd+postcopy/preempt/recovery/tls/psk
> executed in 3.32 secs
> ---
> * Postcopy tests are working well with setting 'multifd = true'.

Great.

> 
> > >         # /x86_64/migration/multifd/tcp/tls/psk/match
> > >         # /x86_64/migration/multifd/tcp/plain/zstd
> > >         # /x86_64/migration/multifd/tcp/plain/cancel
> 
> * Above precopy tests already enable (16) multifd channels and they
> seem to test scenarios like: resume after migrate_cancel() or precopy
> with compression (zstd). Enabling 'postcopy' here is not the same as
> setting 'postcopy=true'. Do we really need to redefine these tests for
> postcopy migration? Does compression (zstd/zlib etc.) OR
> migrate_cancle() work with 'postcopy' migration?

Since multifd doesn't work with postcopy phase, compression so far cannot
happen in postcopy phase but only in precopy phase.

So the tests I suggested was trying to make sure multifd major features (in
this case, tls, compression, and cancellation) work like before even if we
set postcopy-ram=on in the feature list, because after your changes merged,
people may start always set postcopy-ram=on for all cases.

OTOH, these test cases do not test anything that would happen in postcopy
phase, they should be covered by the postcopy tests you added above.

Thanks,

-- 
Peter Xu