[LU-6043] DLC: lnetctl import does not configure any networks

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.7.0
• Fix Version/s: Lustre 2.7.0
• Type: Bug
• Priority: Major
• Reporter: James A Simmons
• Assignee: Amir Shehata (Inactive)
• Resolution: Fixed
• Labels: lnet
• Severity: 3
• Project: Dynamic LNET Configuration

Description

I can bring up a network with lctl net up and once up I used lnetctl export to create a configuration file. After I brought down the network and attempted to bring the network usig lnetctl import < myconfig.yaml it failed with the following message:

>lnetctl import < lnet.yaml
yaml:

• builder:
  errno: -1
  descr: Failed to handle token:10 [state=6, rc=-2]

Issue Links

• is related to: LU-6099 DLC: YAML output must adhere to parsi...
• is related to: LU-6045 DLC: Address a minor inconsistency be...

Comments

[James A Simmons]

Sorry no patch. Not familiar enough with yaml to fix.

[James Nunez (Inactive)]

Amir,

Would you please comment on this ticket?

Thanks,
James

[Gerrit Updater]

Amir Shehata ([e-mail]) uploaded a new patch: http://review.whamcloud.com/13115
Subject: LU-6043 lnet: Fix YAML FSM and other cleanup
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 90204d6df3a9a0385a7401e56d2ca1a0819c0633

[Amir Shehata (Inactive)]

Please note that when you do lnetctl export, you'll get output similar to:

net:
    - net: lo
      nid: 0@lo
      status: up
      tunables:
          peer_timeout: 0
          peer_credits: 0
          peer_buffer_credits: 0
          credits: 0
    - net: tcp
      nid: 192.168.205.156@tcp
      status: up
      interfaces:
          0: eth4
      tunables:
          peer_timeout: 180
          peer_credits: 8
          peer_buffer_credits: 0
          credits: 256
    - net: tcp1
      nid: 192.168.206.167@tcp1
      status: up
      interfaces:
          0: eth5
      tunables:
          peer_timeout: 180
          peer_credits: 8
          peer_buffer_credits: 0
          credits: 256
route:
    - net: tcp3
      gateway: 10.1.1.1@tcp
      hop: 1
      priority: 0
      state: up
    - net: tcp4
      gateway: 10.1.1.1@tcp
      hop: 1
      priority: 0
      state: up

It will dump all currently configured items. To properly test, you'll first need to configure lnet without loading any module parameters. To do that you can issue the following commands:

modprobe lnet
lnetctl lnet configure

Then after you configure lnet:

lnetctl import < config.yaml

When you issue the following command it will attempt to configure each item. If the item is already configured it will fail configuring that particular item (example, if you try to reconfig an already existing network it will fail), but it will not stop and will continue to configure the rest of the items.

Unfortunately, there is a minor inconsistency in the behavior when you add an existing net vs when you add an existing route. The former will error and the latter will be ignored. We can address this inconsistency in a separate patch (but it's a minor issue)

You'll see that when you try to configure the LO Net, since it's part of the configuration but is always configured implicitly and DLC doesn't provide a way to configure the LO Net, it will fail.

the output would likely look similar to

add:
    - net:  // <-- this is the error you get when you try to configure LO
          errno: -2
          descr: "mandatory parameter 'if' not specified. Optionally specify ip2net parameter"
    - net:
          errno: 0
          descr: "success"
    - net:
          errno: 0
          descr: "success"
    - route:
          errno: 0
          descr: "Success"
    - route:
          errno: 0
          descr: "Success"

Currently there is no way to configure LNet from YAML (IE: lnetctl lnet configure). It is assumed that you have already configured LNet either through:

lctl network up
or
lnetctl lnet configure [--all]

If you think it's useful to configure LNet vial YAML, we can address that in a separate patch.

[James A Simmons]

So I did a few different approach with your included patch.

First I found without the patch lnetctl net configure --all only adds lo interface after I run modprobe lnet. The system is configured for o2ib as well. With the patch it does work

The test of running modprobe lnet; lnetctl net configure; lnetctl import < lnet.yaml still fails with the reported error.

[Amir Shehata (Inactive)]

sorry was on vacation.

Can you please share your lnet.yaml file so I can test on my side.

[Gerrit Updater]

Oleg Drokin ([e-mail]) merged in patch http://review.whamcloud.com/13115/
Subject: LU-6043 lnet: Fix YAML FSM and other cleanup
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 943469dc33ab665d3908847a21dfee6a3b574505

[James A Simmons]

Here is the lnet yaml file giving me headaches.

[Amir Shehata (Inactive)]

I fixed your issue in a different patch:
http://review.whamcloud.com/#/c/13304/

[James A Simmons]

You fixed it. Now it works. Thank you.