FR: Ability to make SYNC_COUNT > 1 schedule devices only from same lab |
|||||
Issue descriptionThe autoupdate_P2P test is generally running very well: https://stainless.corp.google.com/search?view=matrix&row=board&col=build&first_date=2018-01-28&last_date=2018-01-31&test=%5Eautoupdate%5C_P2P%24&exclude_cts=true&exclude_not_run=false&exclude_non_release=true&exclude_au=true&exclude_acts=true&exclude_retried=true&exclude_non_production=true However the number 1 failure is "P2P update was disabled because no suitable peer DUT was found." Looking at stainless you can see that some boards pass every time whereas some boards fail a lot. I will investigate what is going on here
,
Jan 31 2018
I think this means the test is being given DUTs in different labs I checked though and chromeos2 and chromeos6 DUTs can ping and ssh into each other. But apparently cannot find each other for P2P updates.
,
Jan 31 2018
Amin, do you know what the criteria is for P2P updates to work between DUTs?
,
Jan 31 2018
I think there is a more info in this README file: https://chromium.git.corp.google.com/chromiumos/platform2/+/e3ac496bc7d22a1d2d9080a3b87458eef3dfcf47/p2p/README.md Can you take a look and see if any information there helps? It seems like the files are advertised through DNS-SD service. Could it be some kind of race condition where a device doesn't find another one because the other is some stale/waiting condition?
,
Feb 1 2018
+infra deputies for clarification on the network setup +pprahbu who did the sync_count magic initially Deputies a couple of questions: Does chromeos2 vs chromeos6 mean that the DUTs are in different labs? So they would be on different LANs and therefore p2p does not work?
,
Feb 1 2018
We currently have no support for targeting multi-DUT tests to the same lab. When the DUTs are determined for incoming multi-DUT test request, *any* two DUTs with matching labels are picked. Since the labels are identical (synch_count just asks for N DUTs with the same labels), these DUTs are always available on a single shard. So far so good. Now if the DUTs on that shard are split between labs, nothing stops those DUTs from being picked together.
,
Feb 1 2018
Re #5: Yep, that reading is correct. Different labs => different vlan => no P2P.
,
Feb 9 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/third_party/autotest/+/620ccb3ee7127557ad61e060819264541443870a commit 620ccb3ee7127557ad61e060819264541443870a Author: David Haddock <dhaddock@chromium.org> Date: Fri Feb 09 07:42:29 2018 Skip autoupdate_P2P when scheduler gives us DUTs from different labs. Also adding in copyright header and renaming control file to be more consistent. BUG= chromium:809681 BUG=chromium:807495 TEST=autoupdate_P2P.local Change-Id: Id67580664331b1d9a6f64eb8600ecbb0b8a3af85 Reviewed-on: https://chromium-review.googlesource.com/905808 Commit-Ready: David Haddock <dhaddock@chromium.org> Tested-by: David Haddock <dhaddock@chromium.org> Reviewed-by: Amin Hassani <ahassani@chromium.org> [modify] https://crrev.com/620ccb3ee7127557ad61e060819264541443870a/server/site_tests/autoupdate_P2P/autoupdate_P2P.py [rename] https://crrev.com/620ccb3ee7127557ad61e060819264541443870a/server/site_tests/autoupdate_P2P/control.delta
,
Feb 9 2018
Turning this into a FR then
,
Feb 9 2018
|
|||||
►
Sign in to add a comment |
|||||
Comment 1 by dhadd...@chromium.org
, Jan 31 2018