Make YARN web interface work with both primary and standby resourcemanager
Open, LowPublic
Actions

Assigned To

None

Authored By

	BTullis
	Mar 7 2023, 3:39 PM

Description

We have a web interface for YARN, which is available at https://yarn.wikimedia.org

However, this only works when an-master1001.eqiad.wmnet is in the active role.

btullis@an-master1001:~$ sudo kerberos-run-command yarn /usr/bin/yarn rmadmin -getServiceState an-master1001-eqiad-wmnet
active

When the active resourcemanager is an-master1002.eqiad.wmnet the web interface automatically tries to redirect the incoming requests to http://an-master1002.eqiad.wmnet:8088

In addition to this, the yarn.wikimedia.org alias only points to a single host, so when this host is unavailable, the YARN web UI is totally unavailable too. It would be nice if this we re highly-available.

Related Objects

Mentioned In: T332573: Refresh an-master100[1-2] with an-master100[3-4]
T331446: Ensure yarn.wikimedia.org point to the active RM

Event Timeline

BTullis created this task.Mar 7 2023, 3:39 PM

Restricted Application added a subscriber: Aklapper. · View Herald TranscriptMar 7 2023, 3:39 PM

• nfraison mentioned this in T331446: Ensure yarn.wikimedia.org point to the active RM.Mar 7 2023, 3:40 PM

FI dns repo:

yarn             1D IN CNAME dyna.wikimedia.org.

With that dyna abstraction to be discussed with traffic team

; "dyna.wikimedia.org." is the focal point of CNAMEs across all our zones for
; most cases that ultimately map to text-addrs.  By using longer-TTL CNAMEs
; pointing to the dynamic short-TTL entry here, we increase DNS cache hitrates
; in the general case across all languages/projects, especially when
; considering edns client subnet fragmentation.  Some exceptions still need
; their own DYNA record rather than a CNAME, such as zone root entries and any
; name which has other metadata alongside its addresses (e.g. MX, SPF, DMARC).
; (www.wikipedia.org was a better/previous choice, but it was moved here due to
; interactions with selective CN censorhsip at the DNS level)
dyna            600 IN DYNA geoip!text-addrs

JArguello-WMF added a project: Shared-Data-Infrastructure.Mar 14 2023, 3:55 PM

JArguello-WMF moved this task from Backlog to To be discussed on the Shared-Data-Infrastructure board.

BTullis added a project: Data-Platform-SRE.Jun 9 2023, 11:56 AM

JArguello-WMF removed a project: Shared-Data-Infrastructure.Jun 29 2023, 1:41 PM

JArguello-WMF removed a project: Data-Engineering-Planning.Jun 29 2023, 9:41 PM

BTullis triaged this task as Low priority.Jul 3 2023, 11:46 AM

Gehel moved this task from Incoming to Misc on the Data-Platform-SRE board.Dec 7 2023, 2:21 PM

BTullis mentioned this in T332573: Refresh an-master100[1-2] with an-master100[3-4].Jan 15 2024, 12:19 PM

Make YARN web interface work with both primary and standby resourcemanagerOpen, LowPublicActions

Description

Related Objects

Event Timeline

Make YARN web interface work with both primary and standby resourcemanager
Open, LowPublic
Actions