#fdio-meeting: FD.io CSIT Project Meeting

Meeting summary

1. Agenda bashing (mackonstan, 13:59:21)
   1. Tibor Frank (tifrank, 13:59:31)
   2. Vratko Polak. (vrpolak, 14:01:28)
   
   
2. VPP inputs (mackonstan, 14:03:25)
   1. Andrew - vpp v23.10 is a thing, will prepare rls schedule, should be ready by next week (mackonstan, 14:03:53)
   2. Andrew - yday vpp-dev call discussions (mackonstan, 14:04:26)
   3. Andrew - no single reference for standard mtu definition incl. jumbo mtu, there is 'max frame size' and there is 'mtu' (mackonstan, 14:06:02)
   4. Andrew - use 'show hardware interface' to see max frame size (mackonstan, 14:06:25)
   5. Andrew - what about virtual interfaces - it depends if chained buffers are supported or not - need to contact maintainers for details (mackonstan, 14:07:13)
   6. Andrew - commenting on Vratko's vpp API msgs versioning email on vpp-dev, csit-dev - Andrew working with Vratko to update msg status (mackonstan, 14:09:04)
   7. Vratko - re mtu/mfs - have some more insight into where things are in VPP, need to work through code and maintainers to work it all out (mackonstan, 14:09:59)
   8. Vratko - Damjan confirmed af_xdp will never work with jumbo as it doesn't support chained buffer; avf does not work with jumbo today due to the way CSIT initializes it (mackonstan, 14:11:36)
   9. Vratko + Maciek - let's work through the latest information, and aim to provide a consistent documentation of current behaviour of VPP and CSIT, and use that as a reference to work with the community on any plan to address mtu/mfs larger then 1500/1518 (mackonstan, 14:12:49)
   
   
3. TSC (mackonstan, 14:13:28)
   1. Dave - re new equipment - no update on 4 SPR ETA, waiting for LFN finance to approve ICX-Ds, waiting for Vexxhost quote for Ampere servers to replace 1n-tx2 (mackonstan, 14:14:29)
   
   
4. Tests - HW accelerators (mackonstan, 14:16:30)
   1. Peter - reported over email issues with QAT2.0, DSA, GTPU/e810 - see email to Kai, Yulong, Dave, Maciek (mackonstan, 14:17:06)
   2. QAT2.0: AT Driver installed and is not working on 3nb-spr, 2n-spr, 3n-snr; Driver failed to install on 3na-spr which results into Ubuntu 22.04 not booting; QAT cannot be initialized within the test. (mackonstan, 14:17:57)
   3. DSA: DSA driver was installed on all capable servers (via Ansible); Changed "intel_iommu=on" to "intel_iommu=on,sm_on" in kernel command line (via Ansible); CSIT lab do not have BIOS option “Opt-Out Illegal MSI Mitigation=Enabled”. We waited for SuperMicro vendor to release BIOS 1.1. Due to release, the upgrade was rescheduled, currently no ETA. BIOS is released. (mackonstan, 14:18:35)
   4. GTPU: Failed to enable GTPU; Not clear what needs to be done CSIT side; Intel to check what might be the root cause behind VPP call. (mackonstan, 14:19:33)
   5. Kai - will lead the resolution of all above from the Intel side, will connect with usual Intel contacts to have them addressed, will report back. (mackonstan, 14:25:15)
   
   
5. Previous Release CSIT-2306 (mackonstan, 14:29:32)
   1. performance RCAs (mackonstan, 14:30:45)
   2. Vratko - updated few jira tickets ... still some work to fully qualify the RCAs (mackonstan, 14:32:27)
   3. Vratko - reported 2 of them on vpp-dev call: wireguard CSIT-1918, ip fragmentation issue VPP-2077. (mackonstan, 14:35:57)
   4. Vratko - following up on above (mackonstan, 14:36:07)
   
   
6. Physical Infrastructure (mackonstan, 14:40:27)
   1. Maciek Peter - SPR mb RMA - still no ETA (mackonstan, 14:40:40)
   
   
7. Virtual Infrastructure, CI/CD (mackonstan, 14:41:30)
   1. Vratko - Dave + Vanessa - planning to do some upgrades to CI/CD tooling, will do in sandbox first, and then schedule in production (mackonstan, 14:42:04)
   
   
8. Trending and Test Failures (mackonstan, 14:42:32)
   1. Vratko - wiki summary to be updated (mackonstan, 14:43:36)

Action items

1. (none)

People present (lines said)

1. mackonstan (33)
2. collab-meetbot (9)
3. vrpolak (1)
4. tifrank (1)
5. dwallacelf (0)
6. jlinkes (0)