Set UCT parameters from UCP API #7019

changchengx · 2021-06-29T02:21:39Z

What

cache UCT configurations in ucp_config_modify API, apply UCT configurations when creating ucp_context and ucp_worker
print UCT configurations in ucp_config_print API
print unused cached UCT configurations in ucp_warn_unused_uct_config API

Why ?

Feture requirement

How ?

cache input key/value in the ucp_config object
copy the cached key/value into ucp_context, apply them to UCT configurations when creating ucp_context
apply the cached key/value into UCT configurations when creating ucp_worker

swx-jenkins3 · 2021-06-29T02:24:18Z

Can one of the admins verify this patch?

changchengx · 2021-07-13T15:50:06Z

It seems there's something wrong with the CI system. I checked some other PRs, they also meet with the same failure case.

src/ucp/core/ucp_context.c

src/ucs/config/parser.c

src/ucp/core/ucp_context.c

src/ucs/config/parser.c

src/ucp/core/ucp_context.c

src/uct/base/uct_md.c

src/ucs/config/parser.h

src/ucp/core/ucp_context.c

changchengx · 2021-07-27T05:53:23Z

There's one conflict here. Shoud I rebase the old used master branch to the current master branch?

changchengx · 2021-07-29T13:21:45Z

Confirmed by Bin & Yossi: we don't need to check the key

leibin2014 · 2021-07-29T13:43:29Z

Confirmed by Bin & Yossi: we don't need to check the key

No, actually we checked, we just don't need to double check here.

changchengx · 2021-08-03T23:59:03Z

@leibin2014
I'm not sure why there're 5 cancelled CI test cases. There's no obvious log shows the failure reason.

src/ucp/core/ucp_context.c

src/uct/base/uct_md.c

src/ucp/core/ucp_context.c

leibin2014

Some more minor comments. Otherwise is ok for me.
@yosefe Could you please start review?

src/ucp/core/ucp_context.c

src/ucp/core/ucp_context.h

src/uct/api/uct.h

changchengx · 2021-09-08T00:55:00Z

@leibin2014 Please help trigger test

leibin2014 · 2021-09-08T01:15:48Z

/azp run

azure-pipelines · 2021-09-08T01:16:09Z

Azure Pipelines successfully started running 3 pipeline(s).

changchengx · 2021-09-09T00:37:18Z

@yosefe I've changed the code according to your comments. I checked the failed CI cases, it seems that they're not related with this PR. Some other PRs also failed at these CI cases around these days.

changchengx · 2021-09-10T02:25:45Z

@yosefe Pass all CI test cases.

changchengx · 2021-09-13T07:48:35Z

It seems that below error is not related with this PR
https://dev.azure.com/ucfconsort/ucx/_build/results?buildId=25219&view=logs&jobId=388a871c-1c79-55fb-9b87-030460a06453

2021-09-13T02:55:09.0312560Z [1631501708.994886] [cb23a5575958:25813:0]       ib_device.c:1332 UCX  ERROR   ibv_create_ah(dlid=49152 sl=0 port=1 src_path_bits=0 dgid=fe80::bace:f6ff:fe0a:11f8 sgid_index=0 traffic_class=0) for UD mlx5 connect on mlx5_bond_1 failed: No such device
2021-09-13T02:55:09.0313481Z [cb23a5575958:25813:0:25813]    ud_iface.c:53   Fatal: iface 0x2505300: failed to get peer address

changchengx · 2021-09-13T07:51:21Z

@leibin2014 Please help trigger re-run below CI case:
https://dev.azure.com/ucfconsort/ucx/_build/results?buildId=25219&view=results

changchengx · 2021-09-13T09:58:22Z

I finished code rebase because there's conflict with master branch.
All CI test cases have been passed.

src/ucp/core/ucp_context.c

src/ucp/core/ucp_context.h

src/ucs/config/parser.h

test/gtest/ucp/ucp_test.cc

changchengx · 2021-09-19T02:01:26Z

@yosefe I've updated the PR according to your 2nd round code-review. All CI test cases have been passed.

src/ucp/api/ucp.h

src/ucp/core/ucp_context.c

src/ucp/core/ucp_worker.c

src/ucs/config/parser.h

src/ucp/core/ucp_worker.c

changchengx · 2021-09-22T07:51:58Z

@leibin2014 Please help trigger to re-run below test:
https://dev.azure.com/ucfconsort/ucx/_build/results?buildId=26296&view=results

changchengx · 2021-09-22T12:28:57Z

@yosefe All CR comments have been fixed and CI test cases have been passed.

yosefe

pls squash without rebase

Signed-off-by: Changcheng Liu <jerrliu@nvidia.com>

Backport below PR from master to integration3 branch openucx#7019 Signed-off-by: Changcheng Liu <jerrliu@nvidia.com>

changchengx changed the title ~~Set UCT parameters from UCP API~~ [DNM/WIP]Set UCT parameters from UCP API Jun 29, 2021

changchengx force-pushed the set_uct_env branch 2 times, most recently from da464be to ecf088d Compare July 1, 2021 08:11

changchengx force-pushed the set_uct_env branch 2 times, most recently from 5d802ee to d85f2e9 Compare July 13, 2021 06:52

changchengx changed the title ~~[DNM/WIP]Set UCT parameters from UCP API~~ Set UCT parameters from UCP API Jul 15, 2021

changchengx force-pushed the set_uct_env branch from d85f2e9 to b8930a6 Compare July 15, 2021 07:49

leibin2014 reviewed Jul 21, 2021

View reviewed changes

leibin2014 reviewed Jul 26, 2021

View reviewed changes

leibin2014 reviewed Aug 5, 2021

View reviewed changes

changchengx force-pushed the set_uct_env branch 3 times, most recently from c73237c to 555266d Compare August 8, 2021 13:16

leibin2014 reviewed Aug 9, 2021

View reviewed changes

src/ucp/core/ucp_context.c Outdated Show resolved Hide resolved

src/ucp/core/ucp_context.c Outdated Show resolved Hide resolved

src/ucp/core/ucp_context.c Outdated Show resolved Hide resolved

leibin2014 approved these changes Aug 11, 2021

View reviewed changes

src/ucp/core/ucp_context.c Outdated Show resolved Hide resolved

src/ucp/core/ucp_context.c Outdated Show resolved Hide resolved

yosefe reviewed Sep 5, 2021

View reviewed changes

yosefe reviewed Sep 16, 2021

View reviewed changes

yosefe reviewed Sep 20, 2021

View reviewed changes

yosefe reviewed Sep 21, 2021

View reviewed changes

src/ucp/core/ucp_worker.c Show resolved Hide resolved

src/ucp/core/ucp_worker.c Outdated Show resolved Hide resolved

src/ucp/core/ucp_worker.c Show resolved Hide resolved

yosefe approved these changes Sep 22, 2021

View reviewed changes

UCP: Set UCT parameters from UCP API

7a2c75a

Signed-off-by: Changcheng Liu <jerrliu@nvidia.com>

changchengx force-pushed the set_uct_env branch from 2e49ff6 to 7a2c75a Compare September 23, 2021 00:43

yosefe merged commit b6197c8 into openucx:master Sep 24, 2021

changchengx deleted the set_uct_env branch September 25, 2021 14:42

changchengx added a commit to changchengx/ucx that referenced this pull request Sep 26, 2021

UCP: Set UCT parameters from UCP API

95bc48d

Backport below PR from master to integration3 branch openucx#7019 Signed-off-by: Changcheng Liu <jerrliu@nvidia.com>

changchengx mentioned this pull request Sep 26, 2021

Set UCT parameters from UCP API yosefe/ucx#209

Merged

pentschev mentioned this pull request Oct 8, 2021

ucp_config_modify doesn't return UCS_ERR_NO_ELEM anymore #7519

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Set UCT parameters from UCP API #7019

Set UCT parameters from UCP API #7019

changchengx commented Jun 29, 2021 •

edited

Loading

swx-jenkins3 commented Jun 29, 2021

changchengx commented Jul 13, 2021

changchengx commented Jul 27, 2021

changchengx commented Jul 29, 2021 •

edited

Loading

leibin2014 commented Jul 29, 2021

changchengx commented Aug 3, 2021

leibin2014 left a comment

changchengx commented Sep 8, 2021

leibin2014 commented Sep 8, 2021

azure-pipelines bot commented Sep 8, 2021

changchengx commented Sep 9, 2021

changchengx commented Sep 10, 2021

changchengx commented Sep 13, 2021

changchengx commented Sep 13, 2021

changchengx commented Sep 13, 2021

changchengx commented Sep 19, 2021

changchengx commented Sep 22, 2021

changchengx commented Sep 22, 2021

yosefe left a comment

Set UCT parameters from UCP API #7019

Set UCT parameters from UCP API #7019

Conversation

changchengx commented Jun 29, 2021 • edited Loading

What

Why ?

How ?

swx-jenkins3 commented Jun 29, 2021

changchengx commented Jul 13, 2021

changchengx commented Jul 27, 2021

changchengx commented Jul 29, 2021 • edited Loading

leibin2014 commented Jul 29, 2021

changchengx commented Aug 3, 2021

leibin2014 left a comment

Choose a reason for hiding this comment

changchengx commented Sep 8, 2021

leibin2014 commented Sep 8, 2021

azure-pipelines bot commented Sep 8, 2021

changchengx commented Sep 9, 2021

changchengx commented Sep 10, 2021

changchengx commented Sep 13, 2021

changchengx commented Sep 13, 2021

changchengx commented Sep 13, 2021

changchengx commented Sep 19, 2021

changchengx commented Sep 22, 2021

changchengx commented Sep 22, 2021

yosefe left a comment

Choose a reason for hiding this comment

changchengx commented Jun 29, 2021 •

edited

Loading

changchengx commented Jul 29, 2021 •

edited

Loading