. Good work!
On Tue, Jan 4, 2022 at 3:18 AM Manuel Arostegui <marostegui(a)wikimedia.org> wrote:
Hello,
If you don’t use
tendril.wikimedia.org or
dbtree.wikimedia.org, feel free to ignore this
message.
As of today, tendril is now retired and the main page is replaced with a list of
replacement for different services tendril used to provide:
For checking out our dbtree and replication data:
if you are in the NDA LDAP group, use Orchestrator
otherwise, use the information page on
noc.wikimedia.org. For more detail you can also
check eqiad.json or codfw.json
If you are looking for slow queries log, go to slow queries dashboard using our standard
observability platform (logstash) (NDA required)
Tendril has been a great tool for us during the years, but unfortunately it is impossible
to maintain with modern MariaDB versions (it uses TokuDB, which is no longer available on
MariaDB after 10.1 and needs to be compiled separately) nor its webservice is compatible
with modern php versions. Its database is still running on Stretch and on MariaDB 10.1
(which has not been supported for a year already) and it is having serious scalability
issues. This would unblock us from removing a lot of legacy home-brew craft and replace
them with more modern toolings such as orchestrator.
Orchestrator has been in place for a few months now, and provides us with a great way to
see and (in the future) manage replication topologies. For now we are using it only for
visualization purposes but in the future we’d like it to also help us to handle
replication changes (it can be done from the UI or via CLI) and recover topologies
automatically if they fail and involve masters or intermediate masters.
The slow queries dashboard in logstash offer multiple advantages over tendril. You can
set the threshold to see slow queries that took longer to run. You can filter out code
paths you’re not interested in or zoom in to relevenet code paths. You can limit it to
write queries or read queries only. Also, it provides id of the request making the slow
query, so you can cross check it with the rest of logstash or hadoop to identify
problematic behavior.
If you need it for the transition period, you can still access it in
tendril-legacy.wikimedia.org. But it will be shut down in a month. You can follow the work
of shutting down tendril in
https://phabricator.wikimedia.org/T297605.
Thank you.
_______________________________________________
Ops mailing list -- ops(a)lists.wikimedia.org
To unsubscribe send an email to ops-leave(a)lists.wikimedia.org
_______________________________________________
Ops mailing list -- ops(a)lists.wikimedia.org
To unsubscribe send an email to ops-leave(a)lists.wikimedia.org