#fdio-csit: FD.io CSIT project meetign
Meeting started by mackonstan at 14:00:53 UTC
(full logs).
Meeting summary
-
- Jan Gelety (jgelety,
14:01:21)
- Tibor Frank (tifrank,
14:01:44)
- Agenda bashing (mackonstan, 14:01:55)
- Vratko Polak. (vrpolak,
14:02:04)
- Physical labs, testbeds (mackonstan, 14:02:34)
- Jan/Peter: more issues with SSDs on Skx, even
though FW got update, looks like a new issue. tb31 3n-skx is
affected, currently operational, monitoring WIP. (mackonstan,
14:04:25)
- Inputs from projects (mackonstan, 14:04:36)
- Inputs from projects - VPP (mackonstan, 14:04:45)
- not much new (mackonstan,
14:04:53)
- Inputs from projects - VSAP (mackonstan, 14:05:03)
- no update (mackonstan,
14:05:07)
- Inputs from projects - TSC (mackonstan, 14:05:16)
- Vratko - long discussion about physical
developer meeting - conclusion: VPP will not go there, but instead
do virtual/webinars due to pandemic/economic impact (mackonstan,
14:06:10)
- Inputs from projects - CNCF CNF Testbed (mackonstan, 14:09:03)
- no update (mackonstan,
14:09:07)
- Releases - CSIT-2001 (mackonstan, 14:09:41)
- Skx NDR issue - eno interrupts are spreading
across all cores on TG TRex nodes. (mackonstan,
14:19:14)
- Skx NDR issue - only observed on Skx, not on
Clx. TRex on Skx only, eno are mgmt interfaces could be related to
ssh activity to TRex during test, but again this is not happening on
Clx (same motherboard as Skx, same mgmt interface and driver) or
other servers in CSIT lab. (mackonstan,
14:26:02)
- Releases - CSIT-2005 (mackonstan, 14:27:27)
- Working on optimizing and unifying trending
jobs for daily and weekly execution, and comparisons (mackonstan,
14:28:42)
- Vratko - looking to have bisecting shipped on
MRR and NDRPDR\ (mackonstan,
14:29:28)
- Releases - CSIT-1908.2 (mackonstan, 14:29:47)
- Dave - VPP is looking to release 1908.2 within
the month of April. (mackonstan,
14:32:12)
- Jan/Maciek - CSIT should be done with Skx
1908.2 testing in two weeks time. (mackonstan,
14:33:34)
- Jan - ci-management patch for CSIT 1908.2
branch needs to be merged (mackonstan,
14:36:40)
- Operations (mackonstan, 14:41:10)
- Committers - Maciek has an action to update
yaml file and screen contributors' work for any substantials to
warrant additions to CSIT committer list (mackonstan,
14:42:34)
- VPP code performance (mackonstan, 14:44:17)
- Vratko - one regression in SRv6 most likely due
to the VPP patch introducing API change, could be due to VPP or CSIT
doing things differently, sent emails. (mackonstan,
14:45:32)
- Vratko - several false anomalies due to
environment, some very fresh anomalies, waiiting to see if it's just
a fluke or for real (mackonstan,
14:46:08)
- Jan - ip6scale tests, due to VPP change, Jakub
Grajcar API changes broke VAT that is still used in CSIT for scale
tests. (mackonstan,
14:47:23)
- Peter - vhost - open tickets for vpp, email to
vpp-dev. (mackonstan,
14:47:45)
- Peter - avf - only skx affected, working on
it. (mackonstan,
14:47:55)
- Developments (mackonstan, 14:48:19)
- CSIT - VM VHOST tests with rxq>1 failing
[VPP-1853] (pmikus,
14:48:31)
- CSIT - VM VHOST tests with qemu mrg_rxbuf=off
[VPP-1854] (pmikus,
14:48:52)
- Vratko - faster PAPI for scale tests, using
async calls and faster method to synthetised the data to send and
parse responses - progressing well, expecting slower than VAT, but
much faster than today (e.g. ip4scale1m - VAT O(10sec), PAPI today
O(30min), asyncPAPI O(~3min) (mackonstan,
14:51:34)
- Govind - having issues with moving to -72
kernel for Ubuntu 18.04 LTS, but was successful with -91. Is it okay
to move to -91? (mackonstan,
14:56:25)
- s/-91/-92/ (mackonstan,
14:56:40)
- Peter - preference for -72. (mackonstan,
14:57:37)
- Govind - will retry -72, and report
back. (mackonstan,
14:57:56)
Meeting ended at 14:58:02 UTC
(full logs).
Action items
- (none)
People present (lines said)
- mackonstan (39)
- collabot` (4)
- pmikus (2)
- vrpolak (1)
- jgelety (1)
- tifrank (1)
Generated by MeetBot 0.1.4.