[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16408908#comment-16408908
]
Todd Lipcon commented on KUDU-1762:
---
BTW I should have noted that the above steps require heap sampling
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16331517#comment-16331517
]
Todd Lipcon commented on KUDU-1762:
---
[~cfreely][~dawn110110][~zhqu...@gmail.com] - anything to report
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16225100#comment-16225100
]
Jean-Daniel Cryans commented on KUDU-1762:
--
[~rains-tung] Do you have any evidence of what was
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16224734#comment-16224734
]
rains-tung commented on KUDU-1762:
--
I was wondering if this bug can be fixed in the future?
Here is my
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16129739#comment-16129739
]
Todd Lipcon commented on KUDU-1762:
---
[~cfreely] [~dawn110110] [~zhqu...@gmail.com] any chance any of you
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16015031#comment-16015031
]
Todd Lipcon commented on KUDU-1762:
---
[~cfreely] any chance you can re-test on a recent 'master' build?
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15935505#comment-15935505
]
Todd Lipcon commented on KUDU-1762:
---
Another is that the WAL appender for every tablet holds onto its
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15935495#comment-15935495
]
Todd Lipcon commented on KUDU-1762:
---
Been looking into this today. Major sources of unaccounted memory
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15861567#comment-15861567
]
Todd Lipcon commented on KUDU-1762:
---
The size of the underlying cfile blocks would definitely affect this
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15860989#comment-15860989
]
ZhangZhen commented on KUDU-1762:
-
{code}
void LRUCache::LRU_Append(LRUHandle* e) {
// Make "e" newest
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15859886#comment-15859886
]
Todd Lipcon commented on KUDU-1762:
---
Hm, I think that issue is tracked by KUDU-972. Good find. Want to
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15858062#comment-15858062
]
Fu Lili commented on KUDU-1762:
---
We have done some testing on this issue, and almost certain that it is
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15823538#comment-15823538
]
DawnZhang commented on KUDU-1762:
-
any further instructions on this issue?
[~tlipcon]
> suspected tablet
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15763484#comment-15763484
]
DawnZhang commented on KUDU-1762:
-
do i need to add "mrs_use_codegen=false" on the profiling server?
>
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15763465#comment-15763465
]
DawnZhang commented on KUDU-1762:
-
10GB rss only occurred on the profiling server. (on server01)
with
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15760006#comment-15760006
]
DawnZhang commented on KUDU-1762:
-
--mrs_use_codegen can not prevents the issue. see the picture.
>
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15744704#comment-15744704
]
DawnZhang commented on KUDU-1762:
-
$ sudo cat /sys/kernel/mm/redhat_transparent_hugepage/defrag
always
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15744586#comment-15744586
]
Todd Lipcon commented on KUDU-1762:
---
Hi Dawn. Is the server in question using transparent huge pages? I
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15728193#comment-15728193
]
DawnZhang commented on KUDU-1762:
-
detail memory usage
{code}
Memory usage by subsystem
Id Parent
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15728140#comment-15728140
]
DawnZhang commented on KUDU-1762:
-
/memz here
{code}
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15727932#comment-15727932
]
Todd Lipcon commented on KUDU-1762:
---
Thanks, Dawn. Can you share also the output of /memz on the daemon
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15727882#comment-15727882
]
DawnZhang commented on KUDU-1762:
-
tcmalloc heap profiling files uploaded.
see
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15714492#comment-15714492
]
DawnZhang commented on KUDU-1762:
-
kudu version:
kudu 1.0.1
revision
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15714488#comment-15714488
]
DawnZhang commented on KUDU-1762:
-
i turned on the flags but HEAP_PROFILE_TIME_INTERVAL seems not work.
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15709406#comment-15709406
]
Todd Lipcon commented on KUDU-1762:
---
Would you mind setting up one of your servers to use tcmalloc heap
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15703791#comment-15703791
]
Todd Lipcon commented on KUDU-1762:
---
That's true, but we've been working under the assumption that the
[
https://issues.apache.org/jira/browse/KUDU-1762?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15703497#comment-15703497
]
Adar Dembo commented on KUDU-1762:
--
This is expected behavior: the root memory tracker's memory
27 matches
Mail list logo
