Hi Aleksandr,
We are currently working on end-to-end tracing with jaeger [1].
As well as adding more trace points (e.g. multisite syncing on the RGW [2]).
Other trace mechanisms should be deprecated, and there is no plan to make
any fixes there, but actually to remove them [3].
I hope we could move forward with all of these efforts soon.
Please provide any input on other spans you would like to see in the system.
Best regards,
Yuval
[1]
https://github.com/ceph/ceph/pull/52114
[2]
https://github.com/ceph/ceph/pull/52707
[3]
https://github.com/ceph/ceph/pull/48971
On Mon, Oct 30, 2023 at 9:07 AM Александр Гуркин <a.i.gurkin(a)outlook.com>
wrote:
Hi all.
I've been experimenting with tracing configurations for ceph from the docs
and it seems like it doesn't work as described.
There is an option using jaeger, described in the documentation -
https://docs.ceph.com/en/latest/jaegertracing/#jaeger-distributed-tracing/
.
Unfortunately, at this time there are only a few spans left inside the
traces, and there is no end-to-end tracing between components. This is not
enough to work.
There is also an option using LTTng and zipkin for visualization,
described in the documentation -
https://docs.ceph.com/en/latest/dev/blkin/#tracing-ceph-with-lttng.
When compilation flags are added, the system stops functioning.
After adding -DWITH_LTTNG=ON - a crash occurs while the rados bench is
running.
After adding the -DWITH_BLKIN=ON flag, the cluster cannot create a pool.
When you add the -DWITH_EVENTTRACE=ON flag, the application does not build
at all.
Are there any plans to restore LTTng functionality?
Are there any plans to improve Jaeger tracing?
Is there any recommended way to use tracing in ceph today?
Thanks in advance.
Aleksandr Gurkin
_______________________________________________
Dev mailing list -- dev(a)ceph.io
To unsubscribe send an email to dev-leave(a)ceph.io