[LU-10391] LNET: Support IPv6

• Type: New Feature
• Resolution: Fixed
• Priority: Major
• Fix Version/s: Lustre 2.16.0
• Labels: IPv6 lnet upstream
• Assignee: Neil Brown
• Reporter: Doug Oucharek (Inactive)

Description

This improvement calls for fully supporting IPv6 throughout LNet and the LND's. This will allow Lustre to run in IPV6 networks, in conjunction with related changes to any configuration and PtlRPC changes where NIDs are processed (e.g. config llog records, Imperative Recovery, etc.).

Issue Links

• is duplicated by: LU-16715 Resolve genl_lock might_sleep bug
• is related to: LU-18041 struct obd_uuid is too small for MGC"NID" uuid
• is related to: LU-13642 lnetctl: Allow IP specification
• is related to: LU-16572 YAML configuration issues
• is related to: LU-17457 LNet ip2nets and routes parameters do not work with large NIDs
• is related to: LU-18135 sanity-lnet test_250: test_250 failed with 1
• is related to: LU-15478 Regression in 005bd7075c LU-10391 lnet: Change lnet_send() to take large-addr nids
• is related to: LU-9823 LNet fails to come up when using lctl but works with lnetctl
• is related to: LU-17294 sanity-lnet test_219: timeout on Ubuntu since 2.15.59
• is related to: LU-17367 Failover on master: Invalid NID string
• is related to: LU-10003 lnetctl error "cannot add network: invalid argument"
• is related to: LU-17638 sanity-lnet test_0: Failed to export global yaml 139
• is related to: LU-17700 lnet ping: failed to ping NID: Protocol error
• is related to: LU-17882 failover-part-* lustre-initialization: mkfs.lustre: Invalid NID string 'trevis-70vm7:trevis-70vm8'
• is related to: LU-18012 OST fails to start when MGS is down with IPv6 config
• is related to: LU-14391 Large network routes
• is related to: LU-14477 lnet doesn't work with no IPV6 support in the kernel
• is related to: LU-15508 Lnet IPv6 work exposed memory leak in GSS layer
• is related to: LU-17460 sanity-lnet test_214 causes a netdevice reference leak with IPv6
• is related to: LU-17629 Regressions with `lnetctl ping`
• is related to: LU-17667 Under sanity-lnet while getting IP handle case where interface have more than one IP
• is related to: LU-17867 build (Make) fail for EL9.2 (Passes EL9.1 and EL9.3)
• is related to: LU-17874 lnetctl export reports error when no nets have been added
• is related to: LU-17895 lnetctl import segfault on junk input
• is related to: LU-18019 Need to check FORCE_LARGE_NID in load_modules_local
• is related to: LU-18171 Update sanity-sec test cases for IPv6
• is related to: LU-18013 conf-sanity test_30b fails with large NIDs
• is related to: LU-10360 use Imperative Recovery logs for client->MDT/OST connections
• is related to: LU-13625 socklnd can select the wrong IP address to bind socket to
• is related to: LU-17743 Add Large NID support for ko2iblnd
• is related to: LU-10359 remove NIDs from config llogs
• is related to: LU-16738 Improve mount.lustre with many MGS NIDs
• is related to: LU-17273 Update GSS context init to support IPv6
• is related to: LU-11978 Increase maximum number of configured NIDs per device
• is related to: LU-16822 test IPv6 NIDs in sanity-lnet.sh
• is related to: LU-13306 allow clients to accept mgs_nidtbl_entry with IPv6 NIDs
• is related to: LU-14288 Enhance nodemap ranges to work better with IPv6
• is related to: LU-17479 Update lnetctl to allow cYAML to handle PyYAML format.
• is related to: LU-13307 add LCFG_NODEMAP_ADD_RANGEv6 records for IPv6
• is related to: LU-13308 update changelog_ext_nid to handle IPv6 addresses
• is related to: LU-13340 add LCFG_ADD_UUIDv6 and related commands
• is related to: LU-16823 add LNet and OBD connect flags for IPv6 peers
• is related to: LU-17455 socklnd: ksocklnd-config needs to be able to handle IPv6 addresses
• is related to: LU-17986 Update sanity-lnet test 111 for IPv6
• is related to: LU-8915 Don't use linux list structure as user land arguments for lnet selftest
• is related to: LU-12511 Prepare lustre for adoption into the linux kernel
• is related to: LU-5881 Allow hostnames in NID
• Trackbacks: JIRA and Confluence Upgrade Plan Problem statement Currently our bug tracking system (JIRA) and wiki (Confluence) are running versions that are frighteningly outofdate, to the point that for our upgrade of Confluence,...

Comments

[Andreas Dilger]

There are still two sanity-lnet subtests being excluded from testing that list LU-10391 as the reason:

always_except LU-10391 253 254

If these are passing now that IPv6 is working, they should be removed from always_except, or a new ticket filed to track whatever issue is preventing them from running (if there isn't already one open).

[Peter Jones]

Marking as resolved because basic feature is present in 2.16. There is still more work to come in 2.17 that can be tracked under a different LU

[James A Simmons]

We have the final clean ups for the 2.16 release. More work is needed but can be finished in the 2.17 release.

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/55767/
Subject: LU-10391 obdclass: truncate large uuids
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 367e607de650ad9a98e50b208f65fda5a9fd4d28

[Gerrit Updater]

"Chris Horn <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/55767
Subject: LU-10391 obdclass: truncate large uuids
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: a8048a976ebd93b7be59c42a6dc58e1f5577105b

[Chris Horn]

We need to tweak the use of FORCE_LARGE_NID in test-framework.sh in order for it to work transparently with other test suites. Most test suites start LNet via load_modules_local, but FORCE_LARGE_NID is not checked or enforced on this path. Opened https://jira.whamcloud.com/browse/LU-18019 for this

[Andreas Dilger]

I would think it should be to do this today like:

Test-Parameters: testlist=sanity-lnet env=FORCE_LARGE_NID

If that is passing, then it should be possible to run a second sanity-lnet session with IPv6 interfaces.

[James A Simmons]

With the recent landings and Maloo setup we can now test IPv6 with sanity-lnet.sh. Looking to figure out how to do Lustre IPv6 testing.

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/53732/
Subject: LU-10391 lnet: Fault injection add/del ioctls to netlink
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: f1c66236c6869a83f57d07206d845bba42d7cd9f

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/53003/
Subject: LU-10391 lnet: update Netlink commands functionality
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: 56af81e1aac0ec1f924249db061514d93912482d