scylladb / scylla-doc-issues

Repository for reporting issues about Scylla documentation (Deprecated)
2 stars 6 forks source link

Issue in page Update Topology Strategy From Simple to Network #854

Closed rbino closed 1 year ago

rbino commented 2 years ago

I'd like report an issue in the Update Topology Strategy From Simple to Network page.

Problem

When following the step-by-step procedure in the Nodes Are On Different Racks section, it's impossible to perform the step 3 ("Alter the strategy as detailed above") because cqlsh can't be used since the binary protocol was disabled in step 1.

Suggest a fix

Provide the necessary details to follow the procedure (should the binary protocol be reenabled? should the procedure be performed with something other than cqlsh?)

This was introduced after https://github.com/scylladb/scylla-doc-issues/issues/835 and this question is also present there in the comments.

tzach commented 2 years ago

@rbino good catch Do you have a working procedure?

rbino commented 2 years ago

@tzach to successfully perform the topology update we re-enabled the binary protocol with nodetool enablebinary before step 3, but we're not aware of the implications of this (i.e. if something can go wrong if some data comes in between that and step 3) and we don't know the reason behind having to do nodetool disablebinary in the first place, so I'm not sure the proposed solution is the correct one, someone from the Scylla team should validate it.

tzach commented 2 years ago

I assume we need to stop all traffic since we are doing a full shutdown @vladzcloudius can you please advise?

annastuchlik commented 2 years ago

@vladzcloudius Could you help us with this issue? It's impossible to follow the procedure because cqlsh cannot be used. Could you look at the considerations above by @rbino and advise how I should correct the procedure?

vladzcloudius commented 2 years ago

@vladzcloudius Could you help us with this issue? It's impossible to follow the procedure because cqlsh cannot be used. Could you look at the considerations above by @rbino and advise how I should correct the procedure?

No, @annastuchlik, the procedure can not be followed as it is now due to reasons described by @rbino in the opening message.

A traffic should be stopped in some other way.

annastuchlik commented 2 years ago

@slivne @tzach The issue reported in https://github.com/scylladb/scylla-doc-issues/issues/835 resulted in a documentation update that makes the procedure impossible to use. If you run nodetool disablebinary, cqlsh cannot be used. See also https://github.com/scylladb/scylla-doc-issues/issues/835#issuecomment-1026234870. https://github.com/scylladb/scylla-doc-issues/issues/835 was closed, but the problem, in fact, hasn't been solved.

Could you tell me how the procedure should be updated?

stale[bot] commented 1 year ago

Thanks for reporting. This issue has been automatically marked as stale because it had no activity for the last few months, and will be closed if no further action taken. If the issue is valid, please add a comment to keep it alive!

rbino commented 1 year ago

The issue is still present in the documentation

vladzcloudius commented 1 year ago

@tzach @annastuchlik @mykaul FYI

tzach commented 1 year ago

@vladzcloudius can you please confirm @annastuchlik comments above Is the problem still valid? if yes, please advise on a new procedure

annastuchlik commented 1 year ago

@vladzcloudius This is not an issue I could fix myself. Please see https://github.com/scylladb/scylla-doc-issues/issues/854#issuecomment-1210638825. If the problem is still valid, I need information about a procedure that is working. Could you provide it or refer me to somebody who could?

vladzcloudius commented 1 year ago

@vladzcloudius can you please confirm @annastuchlik comments above Is the problem still valid? if yes, please advise on a new procedure

@tzach confirmed. The issue is still relevant. I'll think of a better procedure.

vladzcloudius commented 1 year ago

@vladzcloudius can you please confirm @annastuchlik comments above Is the problem still valid? if yes, please advise on a new procedure

@tzach confirmed. The issue is still relevant. I'll think of a better procedure.

I think item (1) in the procedure in question should be changed to: "Shut down any traffic to the cluster. Note: a failure to do so can cause in a data loss."

The rest can remain the same.

annastuchlik commented 1 year ago

Fixed with https://github.com/scylladb/scylladb/pull/12439