New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

abstract_replication_strategy: prepare options: warn about replication_factor corner cases #16115

Open

bhalevy wants to merge 2 commits into scylladb:master from bhalevy:alter-keyspace-replication_factor

Member

bhalevy commented Nov 21, 2023

Using the replication_factor option with
NetworkTopologyStrategy can currently lead to surprising, counter-intuitive results.

When creating a keyspace, the option initially sets the replication factor on all DCs.

But when used in ALTER KEYSPACE queries, the legacy replication_factor option is applied only to
DCs added after the keyspace was created, while
existing DCs are unaffected, and if no DCs were
added, the option is silently ignored.

This change detects when the replication_factor
can be applied safely, and otherwise, it prints a warning.

Fixes #8881

bhalevy added 2 commits

November 21, 2023 15:24


          abstract_replication_strategy: prepare options: warn about replicatio…

4bdd398

…n_factor corner cases

Using the replication_factor option with
NetworkTopologyStrategy can currently lead to surprising,
counter-intuitive results.

When creating a keyspace, the option initially sets
the replication factor on all DCs.

But when used in ALTER KEYSPACE queries, the legacy
`replication_factor` option is applied only to
DCs added after the keyspace was created, while
existing DCs are unaffected, and if no DCs were
added, the option is silently ignored.

This change detects when the `replication_factor`
can be applied safely, and otherwise, it prints a warning.

Fixes scylladb#8881

Signed-off-by: Benny Halevy <bhalevy@scylladb.com>


          cql_query_test: test_rf_expand: add more test cases

bcc66c3

Cover a couple more corner cases related to the expansion
of the replication_factor option with NetworkTopologyStrategy.

Signed-off-by: Benny Halevy <bhalevy@scylladb.com>

bhalevy requested a review from nyh

November 21, 2023 13:25

Member Author

bhalevy commented Nov 21, 2023

avikivity reviewed

View reviewed changes

cql3/statements/ks_prop_defs.cc

                   auto it = options.find(ks_prop_defs::REPLICATION_FACTOR_KEY);
                   if (it != options.end()) {
                       // Expand: the user explicitly provided a 'replication_factor'.
                       rf = it->second;
                       options.erase(it);
+                      logger.warn("Using the \"replication_factor\" option is not recommended, but was used for keyspace \"{}\". "
+                              "Use per-datacenter replication factor options instead.", ks_name);

Member

avikivity Nov 21, 2023

Warning in the log is not helpful. It will likely be noticed a long time after the keyspace was used. Better to warn via CQL.

Why is it not recommended across the board? I can understand it in some cases, but here it's too broad.

nyh requested changes

View reviewed changes

cql3/statements/ks_prop_defs.cc

                   auto it = options.find(ks_prop_defs::REPLICATION_FACTOR_KEY);
                   if (it != options.end()) {
                       // Expand: the user explicitly provided a 'replication_factor'.
                       rf = it->second;
                       options.erase(it);
+                      logger.warn("Using the \"replication_factor\" option is not recommended, but was used for keyspace \"{}\". "
+                              "Use per-datacenter replication factor options instead.", ks_name);

Contributor

nyh Nov 21, 2023

I don't think we want to put a warning in this place. You're suggesting that using "replication_factor" not recommended, Why? It's a feature we added deliberately (and I think @vladzcloudius expressed his enthusiasm with this feature in the last discussion), we called it "auto expand", and it's not un-recommended or deprecated, so why say that it is? If you want to un-recommend it, we need to update the documentation (and discuss it with @tzach and @vladzcloudius ).

cql3/statements/ks_prop_defs.cc

+                      // Warn about non-trivial cases
+                      if (new_dc_rfs) {
+                          if (old_dc_rfs) {
+                              logger.warn("The \"replication_factor\": {} option was applied only to newly added datacenters. "

Contributor

nyh Nov 21, 2023

This is not what I meant when I suggested to warn - this was supposed to be a CQL warning, not a logger warning.
I guess we can have both (in case the user does ALTER KEYSPACE through some sort of automation script and not cqlsh), but I think just the logger, as you did here, is not enough. It doesn't tell the right person of the problem, and it doesn't say it immediately... One person types "ALTER KEYSPACE" in cqlsh, another person notices this warning a few days later.

cql3/statements/ks_prop_defs.cc

-                      for (const auto& dc : tm.get_topology().get_datacenters()) {
-                          options.emplace(dc, *rf);
+                      // Warn about non-trivial cases
+                      if (new_dc_rfs) {

Contributor

nyh Nov 21, 2023

I'm confused about the need for this logic, and all the different counters and variables and hash tables you added in this code... Isn't it enough to warn if the loop above did "options.insert(opt)" on any option, which means one DC's RF was kept unchanged?

Contributor

nyh Nov 21, 2023

By the way, I don't see any way to test if this logic is correct without writing a test that runs various cases and checks for the warnings. You can also do it manually in cqlsh and check the behavior, but then we don't have a regression test (and I as a reviewer don't know what you tested).

cql3/statements/ks_prop_defs.cc

+                                  logger.warn("\"replication_factor\": {} was inherited from a previous replication strategy, where it may have been applied globally in the cluster. "
+                                          "It now applies to the following topology: {}. Current replication options are: {}. "
+                                          "Run repair to rebuild the new token replicas and then cleanup to remove the stale replicas",
+                                          *rf, dc_racks, options);

Contributor

nyh Nov 21, 2023

Interesting. Unless I'm misunderstanding something, this warning wasn't part of #8881 or mentioned in the commit message, but a new kind of safety check you are proposing here.
But I think (please correct me if I'm wrong) that this check is not only relevant when you have a global rf - even if you specify a separate RF per data center, when you set RF in a data center which didn't have one, you need to know to do a repair. I think this message should happen in any such case, not only when replication_factor is used.
Anyway, this check probably belongs in a separate patch with its own justification.

Contributor

ptrsmrn commented Nov 21, 2023

I find the explanation of how this feature works way more clear than how it is described in our documentation - https://opensource.docs.scylladb.com/stable/cql/ddl.html#networktopologystrategy - changing the docs would be helpful for future readers.

ptrsmrn self-requested a review

November 21, 2023 15:37

Contributor

scylladb-promoter commented Nov 22, 2023

🟢 CI State: SUCCESS

✅ - Build
✅ - Sanity Tests
✅ - Unit Tests

Build Details:

Build URL: https://jenkins.scylladb.com/job/scylla-master/job/scylla-ci/4815/
Duration: 7 hr 39 min
Builder: i-0b8495149f0117ca1

Contributor

denesb commented Nov 22, 2023

In the past, there was opposition to this from @vladzcloudius. I don't remember the arguments, just that there were some.

Member Author

bhalevy commented Nov 22, 2023

In the past, there was opposition to this from @vladzcloudius. I don't remember the arguments, just that there were some.

opposition to what exactly?
I don't understand the context of your comment.
To adding log warnings or to return a cql warning (that I have no idea yet how to implement it)

Contributor

denesb commented Nov 23, 2023

opposition to what exactly?
I don't understand the context of your comment.
To adding log warnings or to return a cql warning (that I have no idea yet how to implement it)

Sorry. IIRC the opposition was against a PR which wanted to forbid ambiguous changes to keyspace replication. So it doesn't apply here.

Personally, I think a CQL warning is the way to go.

bhalevy mentioned this pull request

Unable to change RF more than once unless I specify the datacenter #12543

Closed

bhalevy mentioned this pull request

tablets: alter keyspace #16723

Merged

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Ignore NTS's replication_factor tag

5210bc9

This patch removes the support for the "wildcard" replication_factor option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't have to know all datacenter names when creating the keyspace, but ALTER KEYSPACE will require that and the user will have to specify the exact change in replication factors they wish to make by explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Ignore NTS's replication_factor tag

35f844a

This patch removes the support for the "wildcard" replication_factor option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't have to know all datacenter names when creating the keyspace, but ALTER KEYSPACE will require that and the user will have to specify the exact change in replication factors they wish to make by explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Ignore NTS's replication_factor tag

This patch removes the support for the "wildcard" replication_factor option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't have to know all datacenter names when creating the keyspace, but ALTER KEYSPACE will require that and the user will have to specify the exact change in replication factors they wish to make by explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Ignore NTS's replication_factor tag

5103d9c

This patch removes the support for the "wildcard" replication_factor option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't have to know all datacenter names when creating the keyspace, but ALTER KEYSPACE will require that and the user will have to specify the exact change in replication factors they wish to make by explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

1778d5c

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

7b4ca10

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

e46c541

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

7c56c76

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

34f82a0

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

7c71e22

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

2cf7da3

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

b2fe1b4

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

e95f011

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

fc57b44

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

096d725

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

65c9932

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

a8ed472

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

d4b2a47

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

2e2194f

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

83f5e76

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

7d48965

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

0c00fb6

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

e7c88eb

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

2c82e27

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

327d631

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

360a501

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

9a84355

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

b875151

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

mergify bot pushed a commit that referenced this pull request


          Reject ALTER with 'replication_factor' tag

bc1f267

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See #8881, #15391, #16115

(cherry picked from commit b875151)

ptrsmrn added a commit to ptrsmrn/scylladb that referenced this pull request


          Reject ALTER with 'replication_factor' tag

8e87f32

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See scylladb#8881, scylladb#15391, scylladb#16115

(cherry picked from commit b875151)

xemul pushed a commit that referenced this pull request


          Reject ALTER with 'replication_factor' tag

1f44281

This patch removes the support for the "wildcard" replication_factor
option for ALTER KEYSPACE when the keyspace supports tablets.
It will still be supported for CREATE KEYSPACE so that a user doesn't
have to know all datacenter names when creating the keyspace,
but ALTER KEYSPACE will require that and the user will have to
specify the exact change in replication factors they wish to make by
explicitly specifying the datacenter names.
Expanding the replication_factor option in the ALTER case is
unintuitive and it's a trap many users fell into.

See #8881, #15391, #16115

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet