[
https://issues.apache.org/jira/browse/HBASE-28125?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ke Han updated HBASE-28125:
---------------------------
Description:
When using HBase cluster 2.4.17, I noticed that the list_quota_table_sizes
sometimes could return a table whose quota is not set.
h1. Reproduce
{_}This bug cannot be reproduced deterministically{_}. The probability for its
manifestation is 1.1% (Keep executing the commands, and it occurs two times out
of 178 repeatedly executions).
Start up HBase 2.4.17 cluster (2.10.2 HDFS, 1 HMaster, 2 RS)
Executing the following commands,
{code:java}
create_namespace 'uuidef8e6005b9e74092927a4b335424f7c5'
create 'uuid80d6aa3495094cd7b4018d3ab3fe9db8', {NAME =>
'uuid07f904a09baf414d903b2818d3091f28', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}, {NAME =>
'uuida60aaa69834e4f0596b5b3b3c12b2cb8', VERSIONS => 4, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}, {NAME =>
'uuid085c0991aa1c4d4ea145767e7e7bf60c', VERSIONS => 4, COMPRESSION => 'GZ',
BLOOMFILTER => 'NONE', IN_MEMORY => 'true'}, {NAME =>
'uuid1d9a2bc405c64708b9e471ae14794741', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}
create 'uuid5cafa12ce5034015bb597428b294a40d', {NAME =>
'uuid7d9efb39ac94472b90dc60ed3723cdf9', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'NONE', IN_MEMORY => 'true'}, {NAME =>
'uuidde273134b6434fc584990554cfa64b10', VERSIONS => 2, COMPRESSION => 'GZ',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}
clone_table_schema 'uuid80d6aa3495094cd7b4018d3ab3fe9db8',
'uuidb8e2393af3314726890b70ef5871a9d0'
drop 'uuid5cafa12ce5034015bb597428b294a40d'
truncate 'uuidb8e2393af3314726890b70ef5871a9d0'
compaction_state 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
truncate_preserve 'uuidb8e2393af3314726890b70ef5871a9d0'
drop 'uuidb8e2393af3314726890b70ef5871a9d0'
truncate 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
alter 'uuid80d6aa3495094cd7b4018d3ab3fe9db8', {NAME =>
'uuida60aaa69834e4f0596b5b3b3c12b2cb8', METHOD => 'delete'}
disable 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
incr 'uuid80d6aa3495094cd7b4018d3ab3fe9db8',
'uuid863efa8e4f1f44888af0ed139effba33',
'uuid085c0991aa1c4d4ea145767e7e7bf60c:NONE', 3
drop 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
wal_roll 'hregion2,16020'
wal_roll 'hregion1,16020'
create 'uuid4323f716aea24b5fa001f0722cdc66f9', {NAME =>
'uuidd64032ff2e7340fb8832a16430fc14c1', VERSIONS => 3, COMPRESSION => 'GZ',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'false'}, {NAME =>
'uuidc4be1958501543ac86661793a4c144cb', VERSIONS => 2, COMPRESSION => 'GZ',
BLOOMFILTER => 'ROW', IN_MEMORY => 'true'}, {NAME =>
'uuid0c7961e7f67a464387f9f5ce428f08d1', VERSIONS => 4, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}, {NAME =>
'uuida9f7a62b76fa4560834cc5789d5abf3a', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}
major_compact 'uuid4323f716aea24b5fa001f0722cdc66f9',
'uuid0c7961e7f67a464387f9f5ce428f08d1'
incr 'uuid4323f716aea24b5fa001f0722cdc66f9',
'uuidffb29730a2fb4f5f8c18f0c1bc254402',
'uuida9f7a62b76fa4560834cc5789d5abf3a:cc', 3
alter 'uuid4323f716aea24b5fa001f0722cdc66f9', {NAME =>
'uuida9f7a62b76fa4560834cc5789d5abf3a', METHOD => 'delete'}
update_config 'hregion1,16020' {code}
Then execute read command in hbase shell
{code:java}
list_quota_table_sizes
TABLE SIZE
uuid4323f716aea24b5fa001f0722cdc66f9 5133
1 row(s)
Took 0.0278 seconds {code}
uuid4323f716aea24b5fa001f0722cdc66f9 is not set in the quota table but it still
appears in in list_quota_table_sizes results also with a strange value: 5133.
h1. Thoughts
The root cause might be related to how *quota* table reacts when a new table is
created. I am still investigating the root cause (Injecting logs to
MasterQuotaManager to understand why it also records this table).
was:
When using HBase cluster 2.4.17, I noticed that the list_quota_table_sizes
sometimes could return a table whose quota is not set.
h1. Reproduce
{_}This bug cannot be reproduced deterministically{_}. The probability for its
manifestation is 1.1% (Keep executing the commands, and it occurs two times out
of 178 repeatedly executions).
Start up HBase 2.4.17 cluster (2.10.2 HDFS, 1 HMaster, 2 RS)
Executing the following commands,
{code:java}
create_namespace 'uuidef8e6005b9e74092927a4b335424f7c5'
create 'uuid80d6aa3495094cd7b4018d3ab3fe9db8', {NAME =>
'uuid07f904a09baf414d903b2818d3091f28', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}, {NAME =>
'uuida60aaa69834e4f0596b5b3b3c12b2cb8', VERSIONS => 4, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}, {NAME =>
'uuid085c0991aa1c4d4ea145767e7e7bf60c', VERSIONS => 4, COMPRESSION => 'GZ',
BLOOMFILTER => 'NONE', IN_MEMORY => 'true'}, {NAME =>
'uuid1d9a2bc405c64708b9e471ae14794741', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}
create 'uuid5cafa12ce5034015bb597428b294a40d', {NAME =>
'uuid7d9efb39ac94472b90dc60ed3723cdf9', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'NONE', IN_MEMORY => 'true'}, {NAME =>
'uuidde273134b6434fc584990554cfa64b10', VERSIONS => 2, COMPRESSION => 'GZ',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}
clone_table_schema 'uuid80d6aa3495094cd7b4018d3ab3fe9db8',
'uuidb8e2393af3314726890b70ef5871a9d0'
drop 'uuid5cafa12ce5034015bb597428b294a40d'
truncate 'uuidb8e2393af3314726890b70ef5871a9d0'
compaction_state 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
truncate_preserve 'uuidb8e2393af3314726890b70ef5871a9d0'
drop 'uuidb8e2393af3314726890b70ef5871a9d0'
truncate 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
alter 'uuid80d6aa3495094cd7b4018d3ab3fe9db8', {NAME =>
'uuida60aaa69834e4f0596b5b3b3c12b2cb8', METHOD => 'delete'}
disable 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
incr 'uuid80d6aa3495094cd7b4018d3ab3fe9db8',
'uuid863efa8e4f1f44888af0ed139effba33',
'uuid085c0991aa1c4d4ea145767e7e7bf60c:NONE', 3
drop 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
wal_roll 'hregion2,16020'
wal_roll 'hregion1,16020'
create 'uuid4323f716aea24b5fa001f0722cdc66f9', {NAME =>
'uuidd64032ff2e7340fb8832a16430fc14c1', VERSIONS => 3, COMPRESSION => 'GZ',
BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'false'}, {NAME =>
'uuidc4be1958501543ac86661793a4c144cb', VERSIONS => 2, COMPRESSION => 'GZ',
BLOOMFILTER => 'ROW', IN_MEMORY => 'true'}, {NAME =>
'uuid0c7961e7f67a464387f9f5ce428f08d1', VERSIONS => 4, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}, {NAME =>
'uuida9f7a62b76fa4560834cc5789d5abf3a', VERSIONS => 2, COMPRESSION => 'NONE',
BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}
major_compact 'uuid4323f716aea24b5fa001f0722cdc66f9',
'uuid0c7961e7f67a464387f9f5ce428f08d1'
incr 'uuid4323f716aea24b5fa001f0722cdc66f9',
'uuidffb29730a2fb4f5f8c18f0c1bc254402',
'uuida9f7a62b76fa4560834cc5789d5abf3a:cc', 3
alter 'uuid4323f716aea24b5fa001f0722cdc66f9', {NAME =>
'uuida9f7a62b76fa4560834cc5789d5abf3a', METHOD => 'delete'}
update_config 'hregion1,16020' {code}
Then execute read command in hbase shell
{code:java}
list_quota_table_sizes
TABLE SIZE
uuid4323f716aea24b5fa001f0722cdc66f9 5133
1 row(s)
Took 0.0278 seconds {code}
uuid4323f716aea24b5fa001f0722cdc66f9 is not set in the quota table but it still
appears in in list_quota_table_sizes results also with a strange value: 5133.
h1. Thoughts
The root cause might be related to how *quota* table reacts when a new table is
created. I am still investigating the root cause (Injecting logs to
MasterQuotaManager to understand why it also records this table).
Is this normal behavior for list_quota_table_sizes? If this is abnormal
behavior, I can give it a try for fixing.
> list_quota_table_sizes returns a table whose quota is not set
> -------------------------------------------------------------
>
> Key: HBASE-28125
> URL: https://issues.apache.org/jira/browse/HBASE-28125
> Project: HBase
> Issue Type: Bug
> Components: Quotas
> Affects Versions: 2.4.17
> Reporter: Ke Han
> Priority: Major
>
> When using HBase cluster 2.4.17, I noticed that the list_quota_table_sizes
> sometimes could return a table whose quota is not set.
> h1. Reproduce
> {_}This bug cannot be reproduced deterministically{_}. The probability for
> its manifestation is 1.1% (Keep executing the commands, and it occurs two
> times out of 178 repeatedly executions).
> Start up HBase 2.4.17 cluster (2.10.2 HDFS, 1 HMaster, 2 RS)
> Executing the following commands,
> {code:java}
> create_namespace 'uuidef8e6005b9e74092927a4b335424f7c5'
> create 'uuid80d6aa3495094cd7b4018d3ab3fe9db8', {NAME =>
> 'uuid07f904a09baf414d903b2818d3091f28', VERSIONS => 2, COMPRESSION => 'NONE',
> BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}, {NAME =>
> 'uuida60aaa69834e4f0596b5b3b3c12b2cb8', VERSIONS => 4, COMPRESSION => 'NONE',
> BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}, {NAME =>
> 'uuid085c0991aa1c4d4ea145767e7e7bf60c', VERSIONS => 4, COMPRESSION => 'GZ',
> BLOOMFILTER => 'NONE', IN_MEMORY => 'true'}, {NAME =>
> 'uuid1d9a2bc405c64708b9e471ae14794741', VERSIONS => 2, COMPRESSION => 'NONE',
> BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}
> create 'uuid5cafa12ce5034015bb597428b294a40d', {NAME =>
> 'uuid7d9efb39ac94472b90dc60ed3723cdf9', VERSIONS => 2, COMPRESSION => 'NONE',
> BLOOMFILTER => 'NONE', IN_MEMORY => 'true'}, {NAME =>
> 'uuidde273134b6434fc584990554cfa64b10', VERSIONS => 2, COMPRESSION => 'GZ',
> BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'true'}
> clone_table_schema 'uuid80d6aa3495094cd7b4018d3ab3fe9db8',
> 'uuidb8e2393af3314726890b70ef5871a9d0'
> drop 'uuid5cafa12ce5034015bb597428b294a40d'
> truncate 'uuidb8e2393af3314726890b70ef5871a9d0'
> compaction_state 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
> truncate_preserve 'uuidb8e2393af3314726890b70ef5871a9d0'
> drop 'uuidb8e2393af3314726890b70ef5871a9d0'
> truncate 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
> alter 'uuid80d6aa3495094cd7b4018d3ab3fe9db8', {NAME =>
> 'uuida60aaa69834e4f0596b5b3b3c12b2cb8', METHOD => 'delete'}
> disable 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
> incr 'uuid80d6aa3495094cd7b4018d3ab3fe9db8',
> 'uuid863efa8e4f1f44888af0ed139effba33',
> 'uuid085c0991aa1c4d4ea145767e7e7bf60c:NONE', 3
> drop 'uuid80d6aa3495094cd7b4018d3ab3fe9db8'
> wal_roll 'hregion2,16020'
> wal_roll 'hregion1,16020'
> create 'uuid4323f716aea24b5fa001f0722cdc66f9', {NAME =>
> 'uuidd64032ff2e7340fb8832a16430fc14c1', VERSIONS => 3, COMPRESSION => 'GZ',
> BLOOMFILTER => 'ROWCOL', IN_MEMORY => 'false'}, {NAME =>
> 'uuidc4be1958501543ac86661793a4c144cb', VERSIONS => 2, COMPRESSION => 'GZ',
> BLOOMFILTER => 'ROW', IN_MEMORY => 'true'}, {NAME =>
> 'uuid0c7961e7f67a464387f9f5ce428f08d1', VERSIONS => 4, COMPRESSION => 'NONE',
> BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}, {NAME =>
> 'uuida9f7a62b76fa4560834cc5789d5abf3a', VERSIONS => 2, COMPRESSION => 'NONE',
> BLOOMFILTER => 'ROW', IN_MEMORY => 'false'}
> major_compact 'uuid4323f716aea24b5fa001f0722cdc66f9',
> 'uuid0c7961e7f67a464387f9f5ce428f08d1'
> incr 'uuid4323f716aea24b5fa001f0722cdc66f9',
> 'uuidffb29730a2fb4f5f8c18f0c1bc254402',
> 'uuida9f7a62b76fa4560834cc5789d5abf3a:cc', 3
> alter 'uuid4323f716aea24b5fa001f0722cdc66f9', {NAME =>
> 'uuida9f7a62b76fa4560834cc5789d5abf3a', METHOD => 'delete'}
> update_config 'hregion1,16020' {code}
> Then execute read command in hbase shell
> {code:java}
> list_quota_table_sizes
> TABLE SIZE
> uuid4323f716aea24b5fa001f0722cdc66f9 5133
> 1 row(s)
> Took 0.0278 seconds {code}
> uuid4323f716aea24b5fa001f0722cdc66f9 is not set in the quota table but it
> still appears in in list_quota_table_sizes results also with a strange value:
> 5133.
> h1. Thoughts
> The root cause might be related to how *quota* table reacts when a new table
> is created. I am still investigating the root cause (Injecting logs to
> MasterQuotaManager to understand why it also records this table).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
