Page MenuHomePhabricator
Create Task
Maniphest T320397

_etcd-client SRV record missing for conftool cluster
Closed, ResolvedPublic
Actions

Description

I'm currently developing the etcd client for the new L4LB. I wanted to avoid hardcoding conftool hostnames by leveraging the DNS descovery features included in the etcd client but I'm getting the following error:

vgutierrez@lvs6001:~$ ./l4lb etcd --domain eqiad.wmnet
2022/10/10 11:16:02 dns lookup errors: lookup _etcd-client-ssl-conftool._tcp.eqiad.wmnet on 10.3.0.1:53: no such host and lookup _etcd-client-conftool._tcp.eqiad.wmnet on 10.3.0.1:53: no such host

Per https://etcd.io/docs/v3.3/op-guide/clustering/#dns-discovery it seems like _etcd-client and _etcd-client-ssl SRV records should be created:

To help clients discover the etcd cluster, the following DNS SRV records are looked up in the listed order:

_etcd-client._tcp.example.com
_etcd-client-ssl._tcp.example.com
If _etcd-client-ssl._tcp.example.com is found, clients will attempt to communicate with the etcd cluster over SSL/TLS.

Details

SubjectRepoBranchLines +/-
etcd: use the v3-style SRV recordoperations/mediawiki-configmaster+1 -1
confd: use the v3 style srv recordsoperations/puppetproduction+5 -1
etcd: add records compatible with the v3 etcd libraryoperations/dnsmaster+24 -3
Customize query in gerrit

Event Timeline

Vgutierrez created this task.Oct 10 2022, 11:26 AM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptOct 10 2022, 11:26 AM
Vgutierrez triaged this task as Medium priority.Oct 10 2022, 11:26 AM
Vgutierrez added a comment.Oct 10 2022, 12:49 PM

Checking the client implementation for go.etcd.io/etcd/client/v2 v2.305.4 it looks like the SRV discoverer share code with v3: https://github.com/etcd-io/etcd/blob/client/v2.305.4/client/v2/discover.go and that's probably why a recent v2 client requires v3 style SRV records.

Joe subscribed.Oct 10 2022, 12:52 PM

The correct domain to test for read-only clients is conftool.eqiad.wmnet, see https://gerrit.wikimedia.org/r/plugins/gitiles/operations/dns/+/refs/heads/master/templates/wmnet#107

Vgutierrez added a comment.Oct 10 2022, 12:57 PM

you're right in that regard:

vgutierrez@lvs6001:~$ ./l4lb etcd --domain conftool.eqiad.wmnet
2022/10/10 12:55:44 dns lookup errors: lookup _etcd-client-ssl._tcp.conftool.eqiad.wmnet on 10.3.0.1:53: no such host and lookup _etcd-client._tcp.conftool.eqiad.wmnet on 10.3.0.1:53: no such host

but the etcd client is still expecting a SRV of the form _etcd-client-ssl._tcp or _etcd-client._tcp

Joe added a comment.Oct 10 2022, 12:59 PM

yeah this changed with v3. The problem is that AIUI confd uses an older version of the library and expects the simpler form we have now.

We can either add a new set of records or see if we can update confd's dependency.

Volans subscribed.Oct 10 2022, 1:36 PM
In T320397#8304869, @Joe wrote:

The correct domain to test for read-only clients is conftool.eqiad.wmnet, see https://gerrit.wikimedia.org/r/plugins/gitiles/operations/dns/+/refs/heads/master/templates/wmnet#107

Did you mean read-write?

Vgutierrez added a comment.Oct 10 2022, 1:57 PM

hmm from the mentioned documentation on the task description:

If etcd is using TLS, the discovery SRV record (e.g. example.com) must be included in the SSL certificate DNS SAN along with the hostname, or clustering will fail with log messages like the following:

[...] rejected connection from "10.0.1.11:53162" (error "remote error: tls: bad certificate", ServerName "example.com")

but:

$ $ openssl s_client -connect conf1009.eqiad.wmnet:4001 2>/dev/null </dev/null |openssl x509 -noout -ext subjectAltName -subject
X509v3 Subject Alternative Name: 
    DNS:conf1007.eqiad.wmnet, DNS:conf1006.eqiad.wmnet, DNS:conf1009.eqiad.wmnet, DNS:conf1008, DNS:conf1004, DNS:conf1009, DNS:etcd-v3.eqiad.wmnet, DNS:conf1008.eqiad.wmnet, DNS:conf1005, DNS:etcd.eqiad.wmnet, DNS:conf1006, DNS:conf1007, DNS:conf1004.eqiad.wmnet, DNS:conf1005.eqiad.wmnet
subject=CN = etcd-v3.eqiad.wmnet

so potentially we would also need to add conftool.eqiad.wmnet to the SAN list of the TLS certificate

gerritbot added a comment.Oct 10 2022, 2:12 PM

Change 841138 had a related patch set uploaded (by Giuseppe Lavagetto; author: Giuseppe Lavagetto):

[operations/dns@master] etcd: add records compatible with the v3 etcd library

https://gerrit.wikimedia.org/r/841138

gerritbot added a project: Patch-For-Review.Oct 10 2022, 2:12 PM
gerritbot added a comment.Oct 10 2022, 2:27 PM

Change 841139 had a related patch set uploaded (by Giuseppe Lavagetto; author: Giuseppe Lavagetto):

[operations/mediawiki-config@master] etcd: use the v3-style SRV record

https://gerrit.wikimedia.org/r/841139

jijiki moved this task from Incoming 🐫 to 🥋Good First Task on the serviceops board.Oct 17 2022, 3:57 PM
gerritbot added a comment.Oct 18 2022, 9:06 AM

Change 843873 had a related patch set uploaded (by Giuseppe Lavagetto; author: Giuseppe Lavagetto):

[operations/puppet@production] confd: use the v3 style srv records

https://gerrit.wikimedia.org/r/843873

Vgutierrez added a comment.Nov 18 2022, 10:33 AM

ping?

gerritbot added a comment.Nov 21 2022, 9:15 AM

Change 841138 merged by Giuseppe Lavagetto:

[operations/dns@master] etcd: add records compatible with the v3 etcd library

https://gerrit.wikimedia.org/r/841138

gerritbot added a comment.Nov 21 2022, 9:44 AM

Change 843873 merged by Giuseppe Lavagetto:

[operations/puppet@production] confd: use the v3 style srv records

https://gerrit.wikimedia.org/r/843873

Vgutierrez closed this task as Resolved.Nov 22 2022, 11:54 AM
Vgutierrez assigned this task to Joe.
vgutierrez@lvs6001:~$ ./liberica etcd --config /home/vgutierrez/config.yaml 
Using config file: /home/vgutierrez/config.yaml
2022/11/22 11:52:15 Spawning Watchers...
2022/11/22 11:52:15 Watching /conftool/v1/pools/drmrs/cache_text/ats-tls
2022/11/22 11:52:15 Watching /conftool/v1/pools/drmrs/ncredir/nginx
2022/11/22 11:52:15 etcd endpoints discovered: [https://conf1009.eqiad.wmnet.:4001 https://conf1007.eqiad.wmnet.:4001 https://conf1008.eqiad.wmnet.:4001]

endpoints are now being discovered as expected. Thanks @Joe

gerritbot added a comment.Jan 3 2023, 2:04 PM

Change 841139 merged by Giuseppe Lavagetto:

[operations/mediawiki-config@master] etcd: use the v3-style SRV record

https://gerrit.wikimedia.org/r/841139

Stashbot added a comment.Jan 3 2023, 2:05 PM

Mentioned in SAL (#wikimedia-operations) [2023-01-03T14:05:36Z] <oblivian@deploy1002> Started scap: Backport for [[gerrit:841139|etcd: use the v3-style SRV record (T320397)]]

Stashbot added a comment.Jan 3 2023, 2:07 PM

Mentioned in SAL (#wikimedia-operations) [2023-01-03T14:07:30Z] <oblivian@deploy1002> oblivian and oblivian: Backport for [[gerrit:841139|etcd: use the v3-style SRV record (T320397)]] synced to the testservers: mwdebug2002.codfw.wmnet, mwdebug1002.eqiad.wmnet, mwdebug1001.eqiad.wmnet, mwdebug2001.codfw.wmnet

Stashbot added a comment.Jan 3 2023, 2:13 PM

Mentioned in SAL (#wikimedia-operations) [2023-01-03T14:13:34Z] <oblivian@deploy1002> Finished scap: Backport for [[gerrit:841139|etcd: use the v3-style SRV record (T320397)]] (duration: 07m 58s)

Content licensed under Creative Commons Attribution-ShareAlike (CC BY-SA) 4.0 unless otherwise noted; code licensed under GNU General Public License (GPL) 2.0 or later and other open source licenses. By using this site, you agree to the Terms of Use, Privacy Policy, and Code of Conduct. · Wikimedia Foundation · Privacy Policy · Code of Conduct · Terms of Use · Disclaimer · CC-BY-SA · GPL · Credits