setting `blacklistTwcsTables: false` does not include twcs tables in subsequent runs

[zeitounator]

Project board link

For different reasons (even though it is not recommended in reaper's documentation) I want to repair some Twcs tables on my cluster. I have read the documentation and changed the following setting in /etc/cassandra-reaper/cassandra-reaper.yaml

blacklistTwcsTables: false

After that, I restarted the cassandra-reaper service on every node in the cluster. I don't think this has any importance (see below) but reaper is installed in sidecar mode. I restarted several times and checked the config files and the path used on the java command line in my process list.

But I still can't get reaper to include the twcs tables for my application keyspace. I have tried to force run an existing schedule, create a new schedule, run a repair manually from the repair section in gui... they never get included.

I'm far from a java expert and I might ignore some functionalities of the libs/frameworks used on the project to load configuration... but my impression is that this setting is never used anywhere in the code:

• I found the setBlacklistTwcsTables function declared inside the ReaperApplicationConfiguration class.
• This function has usages in the project but only in test classes which are always setting the value to true (the default in all example config files)
• Anyhow, it isn't used anywhere in the ReaperApplicationConfigurationBuilder class where I can see other config setters being called

Unless I missed something, it looks like this needs to be fixed but I'm not fluent enough in java to propose a clean PR.

In case I'm wrong, how can I make reaper obey my desired configuration?

Thanks.

[adejanovski]

Hi @zeitounator,
the setter is not supposed to be used anywhere in the code, it's only a utility method for Dropwizard's configuration system.
What you'll be interested in is where this value is read actually: https://github.com/thelastpickle/cassandra-reaper/blob/eclipse-store/src/server/src/main/java/io/cassandrareaper/service/RepairUnitService.java#L152-L163

This is where it checks for the setting to decide whether or not to filter the TWCS tables when creating a repair run.

Sadly I cannot reproduce the case you describe. I've created a cluster locally and created a keyspace with two tables, one with stcs and the other with twcs.
When I start Reaper using blacklistTwcsTables set to false, I see repair runs created both manually and through schedules selecting both tables:

Then setting it to true and restarting Reaper will correctly apply the filter:

I'd need to know which versions of Cassandra and Reaper you're using, along with some screenshots and tables schemas to assess the situation and see if I can reproduce this issue.

[zeitounator]

Hi @adejanovski. Thanks for getting back so quickly. I'm currently off work with very limited connectivity. But I'll be back next week and will have a look at all references and provide all required information. See you.

[zeitounator]

Hi @adejanovski, here is the information requested.

Cassandra version: 3.11.14

Reaper version: 3.3.4

The schema is as follows. Column/table names redacted where necessary and default values removed for legibility. But this is accurately reproducing my environment.

DESC enveloop

CREATE KEYSPACE enveloop WITH replication = {'class': 'NetworkTopologyStrategy', 'my_local_dc': '3'}  AND durable_writes = true;

CREATE TABLE enveloop.t_enveloppe (
    key text PRIMARY KEY,
    "field1" text,
    "field2" text,
    "field3" text,
    "field4" text,
    "field5" text,
    "field6" text,
    "field7" text,
    field8 text
) WITH compaction = {'class': 'org.apache.cassandra.db.compaction.TimeWindowCompactionStrategy', 'compaction_window_size': '2', 'compaction_window_unit': 'DAYS', 'max_threshold': '32', 'min_threshold': '4'};
CREATE INDEX t_enveloppe_field4_idx ON enveloop.t_enveloppe ("field4");
CREATE INDEX t_enveloppe_field5_idx ON enveloop.t_enveloppe ("field5");
CREATE INDEX t_enveloppe_filed1_idx ON enveloop.t_enveloppe ("field1");
CREATE INDEX t_enveloppe_filed6_idx ON enveloop.t_enveloppe ("field6");
CREATE INDEX t_enveloppe_filed7_idx ON enveloop.t_enveloppe ("field7");

CREATE TABLE enveloop.t_migrations_plugin (
    key text PRIMARY KEY,
    "migrationId" text
) WITH compaction = {'class': 'org.apache.cassandra.db.compaction.SizeTieredCompactionStrategy', 'max_threshold': '32', 'min_threshold': '4'};

CREATE TABLE enveloop.t_bordereau_special (
    key text PRIMARY KEY,
    "filed1" text,
    "filed2" text,
    filed3 text,
    "filed4" text,
    "filed5" text,
    filed6 map<text, text>,
    filed7 text,
    filed8 text
) WITH compaction = {'class': 'org.apache.cassandra.db.compaction.TimeWindowCompactionStrategy', 'compaction_window_size': '2', 'compaction_window_unit': 'DAYS', 'max_threshold': '32', 'min_threshold': '4'};

CREATE TABLE enveloop.t_cle_controle (
    key text PRIMARY KEY,
    "field1" text,
    "field1" text,
    "field1" text,
    "field1" text,
    field1 text,
    "field1" text
) WITH compaction = {'class': 'org.apache.cassandra.db.compaction.SizeTieredCompactionStrategy', 'max_threshold': '32', 'min_threshold': '4'};

CREATE TABLE enveloop.t_bordereau (
    key text PRIMARY KEY,
    "filed1" text,
    "filed2" text,
    "filed3" text,
    "filed4" text,
    filed5 map<text, text>
) WITH compaction = {'class': 'org.apache.cassandra.db.compaction.TimeWindowCompactionStrategy', 'compaction_window_size': '2', 'compaction_window_unit': 'DAYS', 'max_threshold': '32', 'min_threshold': '4'};
CREATE INDEX t_bordereau_field1_idx ON enveloop.t_bordereau ("filed1");
CREATE INDEX t_bordereau_filed4_idx ON enveloop.t_bordereau ("filed4");
CREATE INDEX t_bordereau_field3_idx ON enveloop.t_bordereau ("filed3");
CREATE INDEX t_bordereau_filed2_idx ON enveloop.t_bordereau ("filed2");

I've originally started reaper on each of the 12 nodes with the setting to true and made the modification after finding out only 2 tables were repaired in the above keyspace. Since then I restarted reaper everywhere as already reported above but I still have only two tables repaired on each run as you can see on the following history screenshot with names in the hover bublle:

This is the schedule in place for that keyspace:

Thanks for your support.