[
https://issues.apache.org/jira/browse/HBASE-13260?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14520203#comment-14520203
]
Enis Soztutar commented on HBASE-13260:
---------------------------------------
bq. What is this? This sounds like a nice compromise where we get to reuse
existing code.
It was a quick hack to see the perf of pure HLog. It is here:
https://github.com/enis/hbase/blob/hbase-13260-review/hbase-server/src/main/java/org/apache/hadoop/hbase/procedure2/store/FSHLogProcedureStore.java.
Not sure how that plays with the rest of WALProcStore, some more work would be
needed to handle rolling, and WAL deletion. Agreed that it is a valid approach
(see my earlier comments suggesting so) to re-use FSHlog, but that also will
have some unneeded stuff (WALKey, region, table name, cluster-uuids, etc).
bq. Anyway, I have to dig up my notes from that chat and write up something but
we can discuss that around hbasecon or the meetup the day before (it is a long
conversation with drawings and similar).
+1. Lets gather around and do some brainstorming.
bq. (The zk-less assignment moved the assignment-state in META, because meta is
now co-located with the master).
I thought that we abandoned co-location of meta. I think that we should not do
that.
bq. What to do for 1.1 though? (I suggest we just go WAL-store).
My initial reasoning for this was to re-use what we have, and not support an
additional WAL format with it's own fencing mechanism + rolling + disk format,
etc. Adding all of these is just added complexity and needs maintenance. Now
that at least we have fixed the perf issues in the impl, quantified and partly
justified that having a pure WAL format for procs is better for performance,
and the fact that we are unlikely to get that kind of write perf using a single
region I think it is fine to go with the wal based approach. Whether it is
custom WAL or FSHlog is another discussion though.
> Bootstrap Tables for fun and profit
> ------------------------------------
>
> Key: HBASE-13260
> URL: https://issues.apache.org/jira/browse/HBASE-13260
> Project: HBase
> Issue Type: Bug
> Reporter: Enis Soztutar
> Assignee: Enis Soztutar
> Fix For: 2.0.0, 1.1.0
>
> Attachments: hbase-13260_bench.patch, hbase-13260_prototype.patch
>
>
> Over at the ProcV2 discussions(HBASE-12439) and elsewhere I was mentioning an
> idea where we may want to use regular old regions to store/persist some data
> needed for HBase master to operate.
> We regularly use system tables for storing system data. acl, meta, namespace,
> quota are some examples. We also store the table state in meta now. Some data
> is persisted in zk only (replication peers and replication state, etc). We
> are moving away from zk as a permanent storage. As any self-respecting
> database does, we should store almost all of our data in HBase itself.
> However, we have an "availability" dependency between different kinds of
> data. For example all system tables need meta to be assigned first. All
> master operations need ns table to be assigned, etc.
> For at least two types of data, (1) procedure v2 states, (2) RS groups in
> HBASE-6721 we cannot depend on meta being assigned since "assignment" itself
> will depend on accessing this data. The solution in (1) is to implement a
> custom WAL format, and custom recover lease and WAL recovery. The solution in
> (2) is to have the table to store this data, but also cache it in zk for
> bootrapping initial assignments.
> For solving both of the above (and possible future use cases if any), I
> propose we add a "boostrap table" concept, which is:
> - A set of predefined tables hosted in a separate dir in HDFS.
> - A table is only 1 region, not splittable
> - Not assigned through regular assignment
> - Hosted only on 1 server (typically master)
> - Has a dedicated WAL.
> - A service does WAL recovery + fencing for these tables.
> This has the benefit of using a region to keep the data, but frees us to
> re-implement caching and we can use the same WAL / Memstore / Recovery
> mechanisms that are battle-tested.
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
