Hello,

Is there a specific timeline you'd like us to meet with this? Mainly the goal is to understand urgency for prioritization. Thanks!

Hi!

So the idea is we'd like overall for all of our clusters to have IPv6 reachability. This is not terribly urgent, just a state that has remained a long time and we'd like to rectify.

The request is along the lines of, what domain-specific knowledge do we need to support IPv6 on these clusters, if possible - really if anything bad will happen if we add IPv6 DNS

For the above clusters, it looks like some of the other hosts in them already have IPv6 DNS, so they may be trivial. There are cases when changes to configuration or FERM need to be made to support it, which largely is what this task is asking about.

A quick survey of the clusters above:

• centrallog[12]001 - is handling anycast syslog, which appears to be being mediated on ipv4 by bird if I'm reading this right. It doesn't seem like adding the host dns for ipv6 would be a problem because of this.
• graphite - Seems to be only listening on ipv4 for most of its services. There are a lot of things running, and save for a very few are only on ipv4.
• logstash - All services except for envoy appear to be listening on ipv6
• thanos-fe - all services except for envoy appear to be listening on ipv6
• mwlog - The logging services do not appear to be listening on ipv6

Since we're seem OK in the current state of gradually transitioning to AAAA for all newly-provisioned hosts, I think we can wait for "natural" hardware/OS upgrade lifecycle for these hosts to be gone and their replacements to have ipv6 records, thoughts ?

The approach makes sense to me to avoid duplicating efforts.

In T271138#8061928, @Volans wrote:

• kafka-logging*:
  • have the AAAA record: kafka-logging[2001-2003]
  • lack the AAAA record: kafka-logging[1001-1003]

IIRC these were sorted out some time ago, but didn't get logged as done on this task yet. Here's DNS as of today:

kafka-logging1001.eqiad.wmnet has address 10.64.16.205
kafka-logging1001.eqiad.wmnet has IPv6 address 2620:0:861:102:10:64:16:205
kafka-logging1002.eqiad.wmnet has address 10.64.32.142
kafka-logging1002.eqiad.wmnet has IPv6 address 2620:0:861:103:10:64:32:142
kafka-logging1003.eqiad.wmnet has address 10.64.48.66
kafka-logging1003.eqiad.wmnet has IPv6 address 2620:0:861:107:10:64:48:66

kafka-logging2001.codfw.wmnet has address 10.192.0.94
kafka-logging2001.codfw.wmnet has IPv6 address 2620:0:860:101:10:192:0:94
kafka-logging2002.codfw.wmnet has address 10.192.16.50
kafka-logging2002.codfw.wmnet has IPv6 address 2620:0:860:102:10:192:16:50
kafka-logging2003.codfw.wmnet has address 10.192.32.24
kafka-logging2003.codfw.wmnet has IPv6 address 2620:0:860:103:10:192:32:24

logstash-* hosts have been added to netbox and the sre.dns.netbox cookbook has been run.

mwlog* hosts added to netbox and sre.dns.netbox cookbook has been run.

For the record: we had to revert the thanos-fe* v6 records yesterday since those are not v6 ready yet (cfr T317909). Not sure if we want to reopen this or a followup task? Either works!

Change 875902 had a related patch set uploaded (by Giuseppe Lavagetto; author: Giuseppe Lavagetto):

[operations/puppet@production] kafka-logging: we now support ipv6

https://gerrit.wikimedia.org/r/875902

Change 875902 merged by Giuseppe Lavagetto:

[operations/puppet@production] kafka-logging: we now support ipv6

https://gerrit.wikimedia.org/r/875902