Mauro Tridici
2019-Jul-22 07:42 UTC
[Gluster-users] "du" and "df -hT" commands output mismatch
Hi Hari, I hope that the crawl will run at most for a couple of days. Do you know if there is a way to solve the issue definitely ? GlusterFS version is 3.12.14. You can find below some additional info. Volume Name: tier2 Type: Distributed-Disperse Volume ID: a28d88c5-3295-4e35-98d4-210b3af9358c Status: Started Snapshot Count: 0 Number of Bricks: 12 x (4 + 2) = 72 Transport-type: tcp Many thanks, Mauro> On 22 Jul 2019, at 09:16, Hari Gowtham <hgowtham at redhat.com> wrote: > > Hi, > Yes the above mentioned steps are right. > The way to find if the crawl is still happening is to grep for > quota_crawl in the processes that are still running. > # ps aux | grep quota_crawl > As long as this process is alive, the crawl is happening. > > Note: crawl does take a lot of time as well. And it happens twice. > > On Fri, Jul 19, 2019 at 5:42 PM Mauro Tridici <mauro.tridici at cmcc.it> wrote: >> >> Hi Hari, >> >> thank you very much for the fast answer. >> I think that the we will try to solve the issue disabling and enabling quota. >> So, if I understand I have to do the following actions: >> >> - save on my notes the current quota limits; >> - disable quota using "gluster volume quota /tier2 disable? command; >> - wait a while for the crawl (question: how can I understand that crawl is terminated!? how logn should I wait?); >> - enable quota using "gluster volume quota /tier2 enable?; >> - set again the previous quota limits. >> >> Is this correct? >> >> Many thanks for your support, >> Mauro >> >> >> >> On 19 Jul 2019, at 12:48, Hari Gowtham <hgowtham at redhat.com> wrote: >> >> Hi Mauro, >> >> The fsck script is the fastest way to resolve the issue. >> The other way would be to disable quota and once the crawl for disable >> is done, we have to enable and set the limits again. >> In this way, the crawl happens twice and hence its slow. >> >> On Fri, Jul 19, 2019 at 3:27 PM Mauro Tridici <mauro.tridici at cmcc.it> wrote: >> >> >> Dear All, >> >> I?m experiencing again a problem with gluster file system quota. >> The ?df -hT /tier2/CSP/sp1? command output is different from the ?du -ms? command executed against the same folder. >> >> [root at s01 manual]# df -hT /tier2/CSP/sp1 >> Filesystem Type Size Used Avail Use% Mounted on >> s01-stg:tier2 fuse.glusterfs 25T 22T 3.5T 87% /tier2 >> >> [root at s01 sp1]# du -ms /tier2/CSP/sp1 >> 14TB /tier2/CSP/sp1 >> >> In the past, I used successfully the quota_fsck_new-6.py script in order to detect the SIZE_MISMATCH occurrences and fix them. >> Unfortunately, the number of sub-directories and files saved in /tier2/CSP/sp1 grew so much and the list of SIZE_MISMATCH entries is very long. >> >> Is there a faster way to correct the mismatching outputs? >> Could you please help me to solve, if it is possible, this issue? >> >> Thank you in advance, >> Mauro >> >> >> >> >> >> >> >> _______________________________________________ >> Gluster-users mailing list >> Gluster-users at gluster.org >> https://lists.gluster.org/mailman/listinfo/gluster-users >> >> >> >> >> -- >> Regards, >> Hari Gowtham. >> >> >> > > > -- > Regards, > Hari Gowtham.-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20190722/9ede238f/attachment.html>
Hari Gowtham
2019-Jul-22 08:28 UTC
[Gluster-users] "du" and "df -hT" commands output mismatch
As of now we don't have way to solve it indefinitely. There may be a number of ways accounting mismatch can happen. To solve each way, we need to identify how it happened (the IOs that went through, their order and the timing) with this we need to understand what change is necessary and implement that. This has to done every time we come across an issue that can cause accounting mismatch. Most of the changes might affect the performance. That is a down side. And we don't have a way to collect the above necessary information. To do the above requirements, we don't have enough bandwidth. If anyone from the community is interested, they can contribute to it. We are here to help with them out. On Mon, Jul 22, 2019 at 1:12 PM Mauro Tridici <mauro.tridici at cmcc.it> wrote:> > Hi Hari, > > I hope that the crawl will run at most for a couple of days. > Do you know if there is a way to solve the issue definitely ? > > GlusterFS version is 3.12.14. > You can find below some additional info. > > Volume Name: tier2 > Type: Distributed-Disperse > Volume ID: a28d88c5-3295-4e35-98d4-210b3af9358c > Status: Started > Snapshot Count: 0 > Number of Bricks: 12 x (4 + 2) = 72 > Transport-type: tcp > > Many thanks, > Mauro > > On 22 Jul 2019, at 09:16, Hari Gowtham <hgowtham at redhat.com> wrote: > > Hi, > Yes the above mentioned steps are right. > The way to find if the crawl is still happening is to grep for > quota_crawl in the processes that are still running. > # ps aux | grep quota_crawl > As long as this process is alive, the crawl is happening. > > Note: crawl does take a lot of time as well. And it happens twice. > > On Fri, Jul 19, 2019 at 5:42 PM Mauro Tridici <mauro.tridici at cmcc.it> wrote: > > > Hi Hari, > > thank you very much for the fast answer. > I think that the we will try to solve the issue disabling and enabling quota. > So, if I understand I have to do the following actions: > > - save on my notes the current quota limits; > - disable quota using "gluster volume quota /tier2 disable? command; > - wait a while for the crawl (question: how can I understand that crawl is terminated!? how logn should I wait?); > - enable quota using "gluster volume quota /tier2 enable?; > - set again the previous quota limits. > > Is this correct? > > Many thanks for your support, > Mauro > > > > On 19 Jul 2019, at 12:48, Hari Gowtham <hgowtham at redhat.com> wrote: > > Hi Mauro, > > The fsck script is the fastest way to resolve the issue. > The other way would be to disable quota and once the crawl for disable > is done, we have to enable and set the limits again. > In this way, the crawl happens twice and hence its slow. > > On Fri, Jul 19, 2019 at 3:27 PM Mauro Tridici <mauro.tridici at cmcc.it> wrote: > > > Dear All, > > I?m experiencing again a problem with gluster file system quota. > The ?df -hT /tier2/CSP/sp1? command output is different from the ?du -ms? command executed against the same folder. > > [root at s01 manual]# df -hT /tier2/CSP/sp1 > Filesystem Type Size Used Avail Use% Mounted on > s01-stg:tier2 fuse.glusterfs 25T 22T 3.5T 87% /tier2 > > [root at s01 sp1]# du -ms /tier2/CSP/sp1 > 14TB /tier2/CSP/sp1 > > In the past, I used successfully the quota_fsck_new-6.py script in order to detect the SIZE_MISMATCH occurrences and fix them. > Unfortunately, the number of sub-directories and files saved in /tier2/CSP/sp1 grew so much and the list of SIZE_MISMATCH entries is very long. > > Is there a faster way to correct the mismatching outputs? > Could you please help me to solve, if it is possible, this issue? > > Thank you in advance, > Mauro > > > > > > > > _______________________________________________ > Gluster-users mailing list > Gluster-users at gluster.org > https://lists.gluster.org/mailman/listinfo/gluster-users > > > > > -- > Regards, > Hari Gowtham. > > > > > > -- > Regards, > Hari Gowtham. > > >-- Regards, Hari Gowtham.