[
https://issues.apache.org/jira/browse/HBASE-4213?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13087369#comment-13087369
]
stack commented on HBASE-4213:
------------------------------
bq. If number of RS who have processed the schema change is >= number of active
RS at this moment then master presumes that all RS have acknowledged the schema
change and hence goes ahead with delete operation.
If I add ten regionservers to a cluster in between start of a schema change and
before it completes, could 'if (servers != null && servers.size() >= rsCount)'
never trigger?
Just remove this:
{code}
# Table should be disabled
- raise(ArgumentError, "Table #{table_name} is enabled. Disable it first
before altering.") if enabled?(table_name)
+ # raise(ArgumentError, "Table #{table_name} is enabled. Disable it first
before altering.") if enabled?(table_name)
{code}
... including the comment.
You should log the exception... add it as arg to below:
{code}
+ } catch (KeeperException e) {
+ LOG.warn("Instant schema change failed for table " + tableName );
+ }
{code}
You do this:
{code}
+ this.schemaChangeTracker = new MasterSchemaChangeTracker(getZooKeeper(),
+ this.zooKeeper.schemaZNode, this);
{code}
Do you have to pass schema znode? Can you not ask the object returned by
getZooKeeper for schemaZNode so only need to have two args in this method?
Whats this?
{code}
+ public void checkTableModifiable(final byte [] tableName,
+ EventHandler.EventType eventType)
throws IOException {
+ preCheckTableModifiable(tableName);
+ if (!eventType.isSchemaChangeEvent()) {
+ if (!getAssignmentManager().getZKTable().
+ isDisabledTable(Bytes.toString(tableName))) {
+ throw new TableNotDisabledException(tableName);
+ }
+ }
+ }
{code}
Whats the is disabled check for? We need this any more? Oh, I see... its
needed if its NOT a schema change event.
Javadoc says '+ * @return true if region is opened successfully with new
schema changes.' but method returns void.
Does this need to be synchronized (you syncrhonize on it later in a different
method):
+ for (HRegion hRegion : onlineRegionsOfTable) {
If a new region for this table comes in meantime, it'll be ok? It'll find new
schema?
What happens if master decides to balance a region at this time? Or disable
it? While reopenRegions is running?
What Ted says on delete cf.
Can you look in zk or something rather than wait on a timer before moving on?
{code}
+ // Take a mini nap for changes to take effect.
+ try {
+ Thread.sleep(200);
+ } catch (InterruptedException e) {
+ }
{code}
Remove the '*' from log messages. You'll start an arms race where every log
message will try and add a new type of flashing glyph so it can stand out from
the crowd.
This patch looks great.
> Support instant schema updates with out master's intervention (i.e with out
> enable/disable and bulk assign/unassign)
> --------------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-4213
> URL: https://issues.apache.org/jira/browse/HBASE-4213
> Project: HBase
> Issue Type: Improvement
> Reporter: Subbu M Iyer
> Assignee: Subbu M Iyer
> Fix For: 0.92.0
>
> Attachments: HBASE-4213-Instant_schema_change.patch
>
>
> This Jira is a slight variation in approach to what is being done as part of
> https://issues.apache.org/jira/browse/HBASE-1730
> Support instant schema updates such as Modify Table, Add Column, Modify
> Column operations:
> 1. With out enable/disabling the table.
> 2. With out bulk unassign/assign of regions.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
