Normally you're right, this should get retried at the HBase layer and would
be transparent. However as part of PHOENIX-4130, we have the hbase client
only try the write once, so there's no chance to retry. We did that to
avoid tying up rpc handlers on the server.
Instead, we retry the entire Phoe
check out PHOENIX-5109 , it likely fixes your issue.
Unfortunately it's targeted for 4.15.0 which hasn't been released yet.
Maybe you can backport and see if it works for your query.
On Tue, Aug 20, 2019 at 11:38 AM Ankit Singhal
wrote:
> CREATE local INDEX local_c_h_index ON test_phoenix.app (c
@Aleksandr did you delete rows from the data table, or the index table?
The output you're showing says that you have orphaned rows in the index
table - i.e. rows that exist only in the index table and have no
corresponding row in the data table. If you deleted the original rows in
the data table w
? e.g. indexes, stats...
On Wed, Jun 19, 2019 at 6:18 PM Alexander Batyrshin <0x62...@gmail.com>
wrote:
> Is it possible not to full scan table1 for ’table1.col = ?’, but do this
> check only on subset table1.pk IN (…)?
>
> On 19 Jun 2019, at 23:31, Vincent Poon wrote:
>
>
'table1.col = ?' will be a full table scan of table1 unless you have a
secondary index on table.col
Check the explain plan to see if it's working as expected
On Wed, Jun 19, 2019 at 7:43 AM Alexander Batyrshin <0x62...@gmail.com>
wrote:
> Hello,
> We have 2 tables:
>
> Table1 - big one (2000M+ r
is your max_versions set to 1 ? keep_deleted_cells?
On Tue, Jan 29, 2019 at 10:41 AM talluri abhishek
wrote:
> Hi All,
>
> We are seeing a couple of issues on some of our Phoenix tables where the
> size of the tables keep growing 2-3 times after around 2-3 days of
> ingestion and the read perfo
A lot of work is currently going into handling large numbers of views -
splittable syscat, view management, etc... but agree that it's not ideal.
There's currently no built-in way to do what you want AFAIK, but you can
manage the columns yourself in a separate table:
- store them all in a single c
I'm not able to repro this on latest 4.15.0 :
0: jdbc:phoenix:> explain select * from US_POPULATION where city = 'test';
+-+-++--+
|
e 3000-5000 rows/sec for us. Any other reason upsert
> with limit is so slow?
>
> Thanks,
> Shawn
>
> On Wed, Dec 19, 2018 at 5:04 PM Vincent Poon
> wrote:
>
>> Shawn,
>>
>> Took a quick look, I think what is happening is the UPSERT is done
>> serial
at 2:31 PM Vincent Poon wrote:
>
> Shawn, that sounds like a bug, I would file a JIRA.
>
> On Tue, Dec 18, 2018 at 12:33 PM Shawn Li wrote:
>
>> Hi Vincent & William,
>>
>>
>>
>> Below is the explain plan, both are PARALLEL excuted in plan:
>&
|
>
> | CLIENT 27-CHUNK 3600 ROWS 48114000 BYTES PARALLEL 18-WAY ROUND
> ROBIN FULL SCAN OVER table2 |
>
> | SERVER 200 ROW
> LIMIT
> |
>
> | CLIENT 200 ROW LIMIT
>
>
>
>
>
>
>
> Thanks,
>
&
ck to the other question. Can you explain what is underlying Phoenix
>> implementation for "upsert select limit"? Why it is slower than without
>> "limit" when insert a huge number (2m rows) like ""upsert into table2
>> select * from table1 limit 2,000,
e higher memory
>>>> usage is caused by using scan cache and memstore under the higher
>>>> throughput.
>>>>
>>>>
>>>>Jaanai Zhang
>>>>Best regards!
>>>>
>>>>
(
>
> state CHAR(2) NOT NULL,
>
> city VARCHAR,
>
> population BIGINT,
>
> …
>
> CONSTRAINT my_pk PRIMARY KEY (state));
>
>
>
> Thanks,
>
> Shawn
>
>
>
> On Wed, Dec 12, 2018, 13:42 Vincent Poon
>> Fo
For #2, can you provide the table definition and the statement used? e.g.
Is the table immutable, or does it have indexes?
On Tue, Dec 11, 2018 at 6:08 PM Shawn/Xiang Li wrote:
> Hi,
>
>
>
> 1. Want to check what is underlying running for limit clause used
> in the following Upsert statem
The Apache Phoenix team is pleased to announce the immediate availability
of the 4.14.1 patch release. Apache Phoenix enables SQL-based OLTP and
operational analytics for Apache Hadoop using Apache HBase as its backing
store and providing integration with other projects in the Apache ecosystem
such
Indexes in Phoenix should not in theory cause any cluster outage. An index
write failure should just disable the index, not cause a crash.
In practice, there have been some bugs around race conditions, the most
dangerous of which accidentally trigger a KillServerOnFailurePolicy which
then potentia
We do need to update the docs after PHOENIX-3925, which changed the
behavior from 'recommended' to 'mandatory'.
I'll update the docs now.
On Tue, Oct 9, 2018 at 1:08 PM Ankit Singhal
wrote:
> We do not allow atomic upsert and throw the corresponding exception in the
> cases documented under the
We are planning a Phoenix 4.14.1 release which will have this fix
On Wed, Sep 26, 2018 at 3:36 PM Batyrshin Alexander <0x62...@gmail.com>
wrote:
> Thank you. We will try somehow...
> Is there any chance that this fix will be included in next release for
> HBASE-1.4 (not 2.0)?
>
> On 27 Sep 2018,
19 matches
Mail list logo