...
- VSPERF
- None of those folks are on this call this week
- PS: Will need a review from someone at Juniper for this stuff
- Does the config make sense? They may be in a good position to judge this.
- JG: Going through last hoops of Intel legal approval to share perf doc w/the Docs team
- JG would like to create a WG to work through perf bottlenecks
- Forked DPDK in TF
- Affects VSPERF, per Sridar from 2019-05-28 Meeting notes
- Currently builds from a cloned DPDK
- Would like some background on how/why this is forked
- So they're not building against a moving target
- Appears to be a bad habit
- Should build against a specific tag
- Worth prototyping to build against official DPDK.org?
- PS: +1
- RLB: Will be able to do this once the build is on the community CI, but is there any rush on this now?
- JG: Unknown on the timing
- Will be running vrouter from a container, so would be OK for now
- Would get initial numbers
- But it's an optics thing. People aren't going to like that it's building from a fork.
- Also, cleaner and less code to maintain
- Will be running vrouter from a container, so would be OK for now
- SK: Speculation on why things are this way: sanity & regression testing
- Not yet fully automated, so people have to be told which version to use
- Until it's automated, will be able to do things the right way
- PS: Not the version, it's the source of the code; it's a fork not the master
- SK: All part of the same process
- Community repos
- contrail-vnc & contrail-dev-env aren't being used by commercial contrail?
- Add TF folks to these repos?
- VMB: Should get these moved to TF space instead
- Can move these w/o disturbing anything else
- PS will check with SK, then follow up with helpdesk@ to get repos moved
- RLB: This is being driven internally by Juniper folks; it's only partly an LF responsibility right now
- CI is running for further testing according to Randy
- TPC
- See notes from 2019-06-04 Infra WG Notes
- JG: Doh! This meeting actually starts 30 minutes later now
- Need to send updated invite for this (already updated on the calendar)
- Casey sent an updated invitation at 10:30
- JG: common code between kernel and dpdk vrouter can be good and bad
- can kill some of the performance benefits on dpdk vrouter to acheive common code
- SK: contrail-vnc and contrail-dev-env is still used internally used by Juniper.
- SK: will help with respect to reviews.
- SK: More Architects will be joining the TWS call starting next meeting.
Action items
- Randy BiasFollow up with Juniper to get someone to review the VSPERF config
- Prabhjot Singh Sethi Email dev@ & Sukhdev Kapur to learn more about the contrail-* repos
- Prabhjot Singh Sethi Follow up with Alex about the TPC stuff
- Joseph Gasparakis Write Casey to have him send an updated invite for this meeting
- Randy Bias Talk to Carlo Contavalli and Marcus about the DPDK build thing (build against a tag/release instead of a fork)
- Joseph Gasparakis to start email thread on using vhost interface from dpdk instead of vrouter dpdk on dev mailing list
- Randy Bias need to get us list of code reviewers Juniper and non - Juniper
- Sukhdev Kapur to followup with juniper engineering to check with respect to usage of dpdk from dpdk.org