[Gluster-devel] [Gluster-users] Memory leak in glusterfs

Nithya Balachandran nbalacha at redhat.com
Fri Jun 7 03:09:03 UTC 2019


Hi Abhishek,

Please use statedumps taken at intervals to determine where the memory is
increasing. See [1] for details.

Regards,
Nithya

[1] https://docs.gluster.org/en/latest/Troubleshooting/statedump/


On Fri, 7 Jun 2019 at 08:13, ABHISHEK PALIWAL <abhishpaliwal at gmail.com>
wrote:

> Hi Nithya,
>
> We are having the setup where copying the file to and deleting it from
> gluster mount point to update the latest file. We noticed due to this
> having some memory increase in glusterfsd process.
>
> To find the memory leak we are using valgrind but didn't get any help.
>
> That's why contacted to glusterfs community.
>
> Regards,
> Abhishek
>
> On Thu, Jun 6, 2019, 16:08 Nithya Balachandran <nbalacha at redhat.com>
> wrote:
>
>> Hi Abhishek,
>>
>> I am still not clear as to the purpose of the tests. Can you clarify why
>> you are using valgrind and why you think there is a memory leak?
>>
>> Regards,
>> Nithya
>>
>> On Thu, 6 Jun 2019 at 12:09, ABHISHEK PALIWAL <abhishpaliwal at gmail.com>
>> wrote:
>>
>>> Hi Nithya,
>>>
>>> Here is the Setup details and test which we are doing as below:
>>>
>>>
>>> One client, two gluster Server.
>>> The client is writing and deleting one file each 15 minutes by script
>>> test_v4.15.sh.
>>>
>>> IP
>>> Server side:
>>> 128.224.98.157 /gluster/gv0/
>>> 128.224.98.159 /gluster/gv0/
>>>
>>> Client side:
>>> 128.224.98.160 /gluster_mount/
>>>
>>> Server side:
>>> gluster volume create gv0 replica 2 128.224.98.157:/gluster/gv0/
>>> 128.224.98.159:/gluster/gv0/ force
>>> gluster volume start gv0
>>>
>>> root at 128:/tmp/brick/gv0# gluster volume info
>>>
>>> Volume Name: gv0
>>> Type: Replicate
>>> Volume ID: 7105a475-5929-4d60-ba23-be57445d97b5
>>> Status: Started
>>> Snapshot Count: 0
>>> Number of Bricks: 1 x 2 = 2
>>> Transport-type: tcp
>>> Bricks:
>>> Brick1: 128.224.98.157:/gluster/gv0
>>> Brick2: 128.224.98.159:/gluster/gv0
>>> Options Reconfigured:
>>> transport.address-family: inet
>>> nfs.disable: on
>>> performance.client-io-threads: off
>>>
>>> exec script: ./ps_mem.py -p 605 -w 61 > log
>>> root at 128:/# ./ps_mem.py -p 605
>>> Private + Shared = RAM used Program
>>> 23668.0 KiB + 1188.0 KiB = 24856.0 KiB glusterfsd
>>> ---------------------------------
>>> 24856.0 KiB
>>> =================================
>>>
>>>
>>> Client side:
>>> mount -t glusterfs -o acl -o resolve-gids 128.224.98.157:gv0
>>> /gluster_mount
>>>
>>>
>>> We are using the below script write and delete the file.
>>>
>>> *test_v4.15.sh <http://test_v4.15.sh>*
>>>
>>> Also the below script to see the memory increase whihle the script is
>>> above script is running in background.
>>>
>>> *ps_mem.py*
>>>
>>> I am attaching the script files as well as the result got after testing
>>> the scenario.
>>>
>>> On Wed, Jun 5, 2019 at 7:23 PM Nithya Balachandran <nbalacha at redhat.com>
>>> wrote:
>>>
>>>> Hi,
>>>>
>>>> Writing to a volume should not affect glusterd. The stack you have
>>>> shown in the valgrind looks like the memory used to initialise the
>>>> structures glusterd uses and will free only when it is stopped.
>>>>
>>>> Can you provide more details to what it is you are trying to test?
>>>>
>>>> Regards,
>>>> Nithya
>>>>
>>>>
>>>> On Tue, 4 Jun 2019 at 15:41, ABHISHEK PALIWAL <abhishpaliwal at gmail.com>
>>>> wrote:
>>>>
>>>>> Hi Team,
>>>>>
>>>>> Please respond on the issue which I raised.
>>>>>
>>>>> Regards,
>>>>> Abhishek
>>>>>
>>>>> On Fri, May 17, 2019 at 2:46 PM ABHISHEK PALIWAL <
>>>>> abhishpaliwal at gmail.com> wrote:
>>>>>
>>>>>> Anyone please reply....
>>>>>>
>>>>>> On Thu, May 16, 2019, 10:49 ABHISHEK PALIWAL <abhishpaliwal at gmail.com>
>>>>>> wrote:
>>>>>>
>>>>>>> Hi Team,
>>>>>>>
>>>>>>> I upload some valgrind logs from my gluster 5.4 setup. This is
>>>>>>> writing to the volume every 15 minutes. I stopped glusterd and then copy
>>>>>>> away the logs.  The test was running for some simulated days. They are
>>>>>>> zipped in valgrind-54.zip.
>>>>>>>
>>>>>>> Lots of info in valgrind-2730.log. Lots of possibly lost bytes in
>>>>>>> glusterfs and even some definitely lost bytes.
>>>>>>>
>>>>>>> ==2737== 1,572,880 bytes in 1 blocks are possibly lost in loss
>>>>>>> record 391 of 391
>>>>>>> ==2737== at 0x4C29C25: calloc (in
>>>>>>> /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
>>>>>>> ==2737== by 0xA22485E: ??? (in
>>>>>>> /usr/lib64/glusterfs/5.4/xlator/mgmt/glusterd.so)
>>>>>>> ==2737== by 0xA217C94: ??? (in
>>>>>>> /usr/lib64/glusterfs/5.4/xlator/mgmt/glusterd.so)
>>>>>>> ==2737== by 0xA21D9F8: ??? (in
>>>>>>> /usr/lib64/glusterfs/5.4/xlator/mgmt/glusterd.so)
>>>>>>> ==2737== by 0xA21DED9: ??? (in
>>>>>>> /usr/lib64/glusterfs/5.4/xlator/mgmt/glusterd.so)
>>>>>>> ==2737== by 0xA21E685: ??? (in
>>>>>>> /usr/lib64/glusterfs/5.4/xlator/mgmt/glusterd.so)
>>>>>>> ==2737== by 0xA1B9D8C: init (in
>>>>>>> /usr/lib64/glusterfs/5.4/xlator/mgmt/glusterd.so)
>>>>>>> ==2737== by 0x4E511CE: xlator_init (in
>>>>>>> /usr/lib64/libglusterfs.so.0.0.1)
>>>>>>> ==2737== by 0x4E8A2B8: ??? (in /usr/lib64/libglusterfs.so.0.0.1)
>>>>>>> ==2737== by 0x4E8AAB3: glusterfs_graph_activate (in
>>>>>>> /usr/lib64/libglusterfs.so.0.0.1)
>>>>>>> ==2737== by 0x409C35: glusterfs_process_volfp (in
>>>>>>> /usr/sbin/glusterfsd)
>>>>>>> ==2737== by 0x409D99: glusterfs_volumes_init (in
>>>>>>> /usr/sbin/glusterfsd)
>>>>>>> ==2737==
>>>>>>> ==2737== LEAK SUMMARY:
>>>>>>> ==2737== definitely lost: 1,053 bytes in 10 blocks
>>>>>>> ==2737== indirectly lost: 317 bytes in 3 blocks
>>>>>>> ==2737== possibly lost: 2,374,971 bytes in 524 blocks
>>>>>>> ==2737== still reachable: 53,277 bytes in 201 blocks
>>>>>>> ==2737== suppressed: 0 bytes in 0 blocks
>>>>>>>
>>>>>>> --
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Regards
>>>>>>> Abhishek Paliwal
>>>>>>>
>>>>>>
>>>>>
>>>>> --
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> Regards
>>>>> Abhishek Paliwal
>>>>> _______________________________________________
>>>>> Gluster-users mailing list
>>>>> Gluster-users at gluster.org
>>>>> https://lists.gluster.org/mailman/listinfo/gluster-users
>>>>
>>>>
>>>
>>> --
>>>
>>>
>>>
>>>
>>> Regards
>>> Abhishek Paliwal
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-devel/attachments/20190607/b1a19933/attachment.html>


More information about the Gluster-devel mailing list