[
https://issues.apache.org/jira/browse/CASSANDRA-4018?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13282526#comment-13282526
]
Sylvain Lebresne commented on CASSANDRA-4018:
---------------------------------------------
Some remarks on the patches:
* The patch removes the gc_grace (of 3 months) for schema tables, but schema
tables really need their gc_grace.
* The second patch allows uppercase for properties in create table, but
CASSANDRA-4278 has a more generic solution for that. It'd probably be better to
rebase this on top of CASSANDRA-4278 (once it's committed).
* In HHOM, the table definition in the comments don't correspond to the actual
schema. And since the rest seems to expect the schema in the comment, I'm not
sure this work as expected (as a side note, it'd be safer to reuse the actual
CF comparator rather that redefining the compositeType).
* Given that the peers table will later store potentially multiple token per
host, it seems 'peers' is not the best name. Mostly this will be an inverted
index for tokens, maybe 'tokens' or 'tokens_maps' or something like that would
be more suited? But on a more general level, if we had CASSANDRA-3647, we could
imagine to have a peers system table, where the key would be the unique host id
and we would store as info the inet address and the list of token. Just saying.
* In hints, following CASSANDRA-4120, the row key is the host id, not the
token, so we should probably rename it in the schema (and maybe use the UUID
type).
* In SystemTable, when inserting into Peers, it use the colum name "token", but
the table definition use the name "peer" (which is better since it's the inet
address that this column store).
* In SystemTable, when fetching keys in Peers, instead of doing the query
manually and doing a resultify, why not use CQL3 for the query itself?
* In SystemTable.getLocalHostId, the PEERS_CF is queried instead of LOCAL_CF
* In NodeIdCf, the id is actually a UUID (and the code does rely on the sorting
being the one of (time) UUID).
Otherwise, after this patch we will just have IndexCf and NodeIdCf that will
have case sensitive names. Do we want to do something about that? I guess
that's not the most useful table to be queried by users, so probably it's not a
big deal, but wanted to mention it to get other opinions.
> Add column metadata to system columnfamilies
> --------------------------------------------
>
> Key: CASSANDRA-4018
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4018
> Project: Cassandra
> Issue Type: Improvement
> Affects Versions: 1.1.0
> Reporter: Jonathan Ellis
> Assignee: Jonathan Ellis
> Priority: Minor
> Fix For: 1.2
>
>
> CASSANDRA-3792 adds this to the schema CFs; we should modernize the other
> system CFs as well
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
