v1.4 when will be first release?

[interduo]

When will be first release of v1.4?

I am thinking about testing XDP_BRIDGE feature.
btw Do You know how big is performance prize changing XDP_BRIDGE form kernel bridge?

[rchac]

I'm seeing a 30-45% CPU use reduction with v1.4 and XDP_BRIDGE. We need to fix some performance aspects for lqosd to make it really smooth for ISPs >2000 subs. But we are getting closer. We're aiming for a release around March.

[interduo]

Will lqosd be required in 1.4?
All the critical things (without lqosd) are fixed?

[thebracket]

lqosd is required: it manages all of the eBPF (XDP/TC) program functionality. The shaper won't start without it. There's more to do in general, beyond the daemon. Documentation, scheduler, debugging, etc.

…

On Tue, Jan 17, 2023 at 5:59 AM Interduo ***@***.***> wrote: But all of the things instead of lqosd are done allready? Will lqosd be required in 1.4? — Reply to this email directly, view it on GitHub <#203 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADRU432RVRZN5HXV5LM5JQ3WS2CRRANCNFSM6AAAAAAT44QNXY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[dtaht]

In terms of a schedule, we are in "play time", now. I have a hope that we, working together, can close 7 more bugs a week than come in, so we can do a feature freeze in the Feburary 7th timeframe. I have hope that more folk will lean in on the rust, or spin out and add a feature of their own that they want. For example, me being a command line guy, makes me want (someone else!) to work on the lqtop utility to pull many of the same things out of lqosd that the gui presently does. We also need to get someone(s) less color blind than h and I to get the graphs making more sense. I think we've now proven that with what we got, a lower sample rate by default is better (perhaps 100ms? I would be sad if we have to go to 1sec) for those with vm architectures. After that I'd like to be spending more time hardening the system, and beating it up, for perhaps an official beta March 7, and a release March 17th.

In terms of beating it up better, my own efforts to emulate 10k users have stalled out, and naturally, that's where we've run into problems at the moment. Trexx or something lighter weight than the containers I attempted to use to emulate delay - is needed. I hit several bugs in other test tools than flent...

I am presently looking at new feature requests with a jaundiced eye, and punting most of them to the v1.5 release, as tempting as it is to (for example) make lqosd speak the netlink protocol directly.

Presently the influxdb is something of a bastard stepchild. Given that we just proved that it was wildly inaccurate, I am going to recommend that people purge or retire their existing influxdb data upon release.

[interduo]

v1.4-rc9 is out and doing its work pretty well and no new issues/errors. What needs to be done for v1.4 release?

[thebracket]

A little more end-to-end testing on the LTS client, a ton of merging.

…

On Mon, Oct 16, 2023, 5:30 PM Interduo ***@***.***> wrote: v1.4-rc9 is out and doing its work pretty well and no new issues/errors. What needs to be done for v1.4 release? — Reply to this email directly, view it on GitHub <#203 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADRU436ETKJSEG7ZTA6WSMTX7WYOVAVCNFSM6AAAAAAT44QNX2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TEOJXHAYTQ> . You are receiving this because you commented.Message ID: ***@***.***>