:: [devuan-dev] Meeting notes 2021-10-…
トップ ページ
このメッセージを削除
このメッセージに返信
著者: B Stack
日付:  
To: devuan developers internal list
題目: [devuan-dev] Meeting notes 2021-10-21
# Devuan meet 2021-10-21 @20:30 UTC

Present: Hendrik, plasma41, fsr, bb|hcb, golinux, rrq

## Old Business
* Website is starting to look complete. The links seem actually to go
where they need to.
* (gl) It is done thanks to LeePen and rrq
* They did a good job.
*
https://www.devuan.org/os/documentation/install-guides/chimaera/upgrade-to-chimaera
still does not warn about impending wicd disappearance, which can
disable network access (making package installation difficult) for
wicd users if they do not replace it before upgrading. The mention of
wicd in the release notes is not enough.
* (gl) the warning about this is currently in proposed repo.

## Old Actions

## New Business

### LeePen Apologies, I am not able to attend.
* Setup britney to manage daedalus migrations from ceres.
* New wicd transitional package in chimaera-proposed-updates. Designed
to ensure users upgrading from beowulf who have wicd installed don't
end up without a network manager. Posted request for testers to DNG.
* Fixed dak gz index file generation in chimaera-* and daedalus.
* Caught up with packaging in ceres. New packages:
devuan-lintian-profile, pdns-recursor, distro-info-data,
network-manager, pcsc-lite, popularity-contest, reportbug, sysvinit,
rsyslog, udisks2, util-linux. Note that pdns-recursor upstream have
dropped 32-bit support: their y2k38 audit requires sizeof(time_t)==8.
So there are only adm64, arm64 and ppc64el packages. Debian has the
same issue. (hendrik) Haven't the 32-bit C libraries provided a 64-bit
type for time? I thought that was available ages ago.
* Fixed bug in releasebot when an open issue has no assignee.

### bgstack15
* Updated original package systemctl-service-shim
  * systemctl replacement script now controls a symlink,
    "/bin/systemctl" which is hardcoded into certain freeipa commands
    (drat those freeipa authors)
* Updated freeipa, xrdp


### bb|hcb Infra stuff
* new DNS
  * `ns4.devuan.dev` (Sofia, BG)
  * `ns5.devuan.dev` (Mathura, IN)
  * `ns6.devuan.dev` (Saint Petersburg, RU) already active on
    `devuan.(dev|org|net)` to be decided for the rest of Devuan domains
  * added `zonediff.sh` to help identify the differences between zone
    git and reality before overwriting the zones; `vim` users may `vi -d
    abc.zone ../abc.zone` for a bidi merge, other editors may or may not
    have this feature
  * I made a mistake and `devuan.(dev|org|net)` now have serials in the
    future; this will get fixed by itself next month, for now just
    increment the serial if you touch
  * rrq appointed me to take care of DNS
  * setup `devuan.info`
* proposed change to networking
  * full mesh between the nodes with `gretap` tunnels
  * bridged with split horizon (in simple words only the direct tunnel
    carries traffic between the hosts - that is L2 network using the L3
    paths)
  * in networking terms that is MP2MP VPLS L2VPN
  * different internal networks use 802.1q tagged traffic on top of the
    transport
  * currently in use for the DNS internal lan (802.1q vlan 1000)
  * the most visible benefit from this will be in the `gnt_br` network
    that is used for DRBD replication, because this model works in pure
    kernel mode with much higher performance
  * packet fragmentation will always be there (both in current and
    proposed models) because the Internet can only do 1500 MTU packets
    (lowering the internal LANs MTU is worse than the fragmentation)
  * this model can coexist with the current `rrqnet` tunnels (that have
    their benefits like nat traversal, higher flexibility and built-in
    loop protection); e.g. `ns3` (`borta`) currently joins the game via
    `rrqnet` via `nash`
* upgraded `ns1` and `ns2` to chimaera
  * `ns1` went flawless
  * `ns2` needed pushing, pulling and kicking:
    * rm /var/lib/nsd/nsd.db
    * restart nsd
    * bump-up zone serials in `ns1`


### golinux
* submitted 2 corrections to The Register review and both were accepted.
  * 1. In the original headline we were offering "3 bootloaders" instead
    of "3 init systems"
  * 2. The VUA's were given credit for the release despite the fact that
    there was no mention of them in the release announcement. There
    wasn't in the ASCII announcement either. Sheesh . . .
* Tooting my horn a little here. This in a distrowatch review:
  * [chimaera] "looks so gorgeous I almost didn't want to mess about
    customizing it."
* Was going to play hooky but jaromil said he would be here . . .
  * What a tease!  A no-show!! :)


### rrq
  * cleaned up the installer-iso to have a single, "top level" target
    codename configuration so that the same code is used an all (both)
    architectures,
  * started the "daedalus preview" stream of installer isos
  * begun flagging upgrading infrastructure hosts to chimaera; some are
    already there. The main potholes are that python2 goes away, and
    something happens to php, but we might be fine with that.
  * upgrade of the nodes will transition us to ganeti 3.0 and it will be
    a major task that is likely to require some hour total downtime
    (probably less than an hour since it's all going to go very smooth)
  * we are in need of a git store (gitea) sysadmin; I'm starting to get
    bored :)
  * we also need a forum (php) sysadmin; same reason.


## New Actions