[Gluster-users] Trying to fix files that don't want to heal
Hi, I have a distributed dispersed volume with files that don't want to heal. I'm trying to fix them manually. I'm currently working on a file that is present on all bricks, GFID exists in .glusterfs-structure and getfattr shows identical attributes for all files. They look like this: # getfattr -m. -d -e hex $brick/somepath/libssl.so.1.1 getfattr: Removing leading '/' from absolute path names # file: $brick/$somepath/libssl.so.1.1 trusted.ec.config=0x080602000200 trusted.ec.dirty=0x0001 trusted.ec.size=0x000a trusted.ec.version=0x00040005 trusted.gfid=0xdd7dd64f6bb34b5f891a5e32fe83874f trusted.gfid2path.0c3a5b76c518ef60=0x34663064396234332d343730342d343634352d613834342d3338303532336137346632662f6c696273736c2e736f2e312e31 trusted.gfid2path.578ce2ec37aa0f9d=0x31636136323433342d396132642d343039362d616265352d6463353065613131333066632f6c696273736c2e736f2e312e31 trusted.glusterfs.quota.1ca62434-9a2d-4096-abe5-dc50ea1130fc.contri.3=0x00029201 trusted.glusterfs.quota.4f0d9b43-4704-4645-a844-380523a74f2f.contri.3=0x00029201 trusted.pgfid.1ca62434-9a2d-4096-abe5-dc50ea1130fc=0x0001 trusted.pgfid.4f0d9b43-4704-4645-a844-380523a74f2f=0x0001 pgfid is "parent gfid" right? Both GFID's refer to a dir in my volume, both of those dirs contain a file named libssl.so.1.1. They seem to be hardlinks: find $brick/$somepath -samefile $brick/$someotherpath/libssl.so.1.1 $brick/$somepath/libssl.so.1 This exceeds the limits of my GlusterFS knowledge. Is that something that can/should happen? If not, is it the reason that file won't heal and how do I fix that? Kind regards Gudrun Amedick smime.p7s Description: S/MIME cryptographic signature Community Meeting Calendar: APAC Schedule - Every 2nd and 4th Tuesday at 11:30 AM IST Bridge: https://bluejeans.com/441850968 NA/EMEA Schedule - Every 1st and 3rd Tuesday at 01:00 PM EDT Bridge: https://bluejeans.com/441850968 Gluster-users mailing list Gluster-users@gluster.org https://lists.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] Trying to fix files that don't want to heal
Hey Gudrun, Could you please try to use the scripts and try to resolve it. We have written some scripts and it is in final phase to get merge - https://review.gluster.org/#/c/glusterfs/+/23380/ You can find the steps to use these scripts in README.md file --- Ashish - Original Message - From: "Gudrun Mareike Amedick" To: "Gluster-users" Sent: Thursday, November 28, 2019 3:57:18 PM Subject: [Gluster-users] Trying to fix files that don't want to heal Hi, I have a distributed dispersed volume with files that don't want to heal. I'm trying to fix them manually. I'm currently working on a file that is present on all bricks, GFID exists in .glusterfs-structure and getfattr shows identical attributes for all files. They look like this: # getfattr -m. -d -e hex $brick/somepath/libssl.so.1.1 getfattr: Removing leading '/' from absolute path names # file: $brick/$somepath/libssl.so.1.1 trusted.ec.config=0x080602000200 trusted.ec.dirty=0x0001 trusted.ec.size=0x000a trusted.ec.version=0x00040005 trusted.gfid=0xdd7dd64f6bb34b5f891a5e32fe83874f trusted.gfid2path.0c3a5b76c518ef60=0x34663064396234332d343730342d343634352d613834342d3338303532336137346632662f6c696273736c2e736f2e312e31 trusted.gfid2path.578ce2ec37aa0f9d=0x31636136323433342d396132642d343039362d616265352d6463353065613131333066632f6c696273736c2e736f2e312e31 trusted.glusterfs.quota.1ca62434-9a2d-4096-abe5-dc50ea1130fc.contri.3=0x00029201 trusted.glusterfs.quota.4f0d9b43-4704-4645-a844-380523a74f2f.contri.3=0x00029201 trusted.pgfid.1ca62434-9a2d-4096-abe5-dc50ea1130fc=0x0001 trusted.pgfid.4f0d9b43-4704-4645-a844-380523a74f2f=0x0001 pgfid is "parent gfid" right? Both GFID's refer to a dir in my volume, both of those dirs contain a file named libssl.so.1.1. They seem to be hardlinks: find $brick/$somepath -samefile $brick/$someotherpath/libssl.so.1.1 $brick/$somepath/libssl.so.1 This exceeds the limits of my GlusterFS knowledge. Is that something that can/should happen? If not, is it the reason that file won't heal and how do I fix that? Kind regards Gudrun Amedick Community Meeting Calendar: APAC Schedule - Every 2nd and 4th Tuesday at 11:30 AM IST Bridge: https://bluejeans.com/441850968 NA/EMEA Schedule - Every 1st and 3rd Tuesday at 01:00 PM EDT Bridge: https://bluejeans.com/441850968 Gluster-users mailing list Gluster-users@gluster.org https://lists.gluster.org/mailman/listinfo/gluster-users Community Meeting Calendar: APAC Schedule - Every 2nd and 4th Tuesday at 11:30 AM IST Bridge: https://bluejeans.com/441850968 NA/EMEA Schedule - Every 1st and 3rd Tuesday at 01:00 PM EDT Bridge: https://bluejeans.com/441850968 Gluster-users mailing list Gluster-users@gluster.org https://lists.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] Trying to fix files that don't want to heal
Hi Ashish, thanks for your reply. To fulfill the "no IO"-requirement, I'll have to wait until second week of December (9th – 14th). We originally planned to update GlusterFS from 4.1.7 to 5 and then to 6 in December. Should we do that upgrade before or after running those scripts? Kind regards GudrunAm Freitag, den 29.11.2019, 00:38 -0500 schrieb Ashish Pandey: > Hey Gudrun, > > Could you please try to use the scripts and try to resolve it. > We have written some scripts and it is in final phase to get merge - > https://review.gluster.org/#/c/glusterfs/+/23380/ > > You can find the steps to use these scripts in README.md file > > --- > Ashish > > From: "Gudrun Mareike Amedick" > To: "Gluster-users" > Sent: Thursday, November 28, 2019 3:57:18 PM > Subject: [Gluster-users] Trying to fix files that don't want to heal > > Hi, > > I have a distributed dispersed volume with files that don't want to heal. I'm > trying to fix them manually. > > I'm currently working on a file that is present on all bricks, GFID exists in > .glusterfs-structure and getfattr shows identical attributes for all > files. They look like this: > > # getfattr -m. -d -e hex $brick/somepath/libssl.so.1.1 > getfattr: Removing leading '/' from absolute path names > # file: $brick/$somepath/libssl.so.1.1 > trusted.ec.config=0x080602000200 > trusted.ec.dirty=0x0001 > trusted.ec.size=0x000a > trusted.ec.version=0x00040005 > trusted.gfid=0xdd7dd64f6bb34b5f891a5e32fe83874f > trusted.gfid2path.0c3a5b76c518ef60=0x34663064396234332d343730342d343634352d613834342d3338303532336137346632662f6c696273736c2e736f2e312e31 > trusted.gfid2path.578ce2ec37aa0f9d=0x31636136323433342d396132642d343039362d616265352d6463353065613131333066632f6c696273736c2e736f2e312e31 > trusted.glusterfs.quota.1ca62434-9a2d-4096-abe5-dc50ea1130fc.contri.3=0x00029201 > trusted.glusterfs.quota.4f0d9b43-4704-4645-a844-380523a74f2f.contri.3=0x00029201 > trusted.pgfid.1ca62434-9a2d-4096-abe5-dc50ea1130fc=0x0001 > trusted.pgfid.4f0d9b43-4704-4645-a844-380523a74f2f=0x0001 > > pgfid is "parent gfid" right? Both GFID's refer to a dir in my volume, both > of those dirs contain a file named libssl.so.1.1. They seem to be > hardlinks: > > find $brick/$somepath -samefile $brick/$someotherpath/libssl.so.1.1 > $brick/$somepath/libssl.so.1 > > This exceeds the limits of my GlusterFS knowledge. Is that something that > can/should happen? If not, is it the reason that file won't heal and how > do > I fix that? > > Kind regards > > Gudrun Amedick > > > Community Meeting Calendar: > > APAC Schedule - > Every 2nd and 4th Tuesday at 11:30 AM IST > Bridge: https://bluejeans.com/441850968 > > NA/EMEA Schedule - > Every 1st and 3rd Tuesday at 01:00 PM EDT > Bridge: https://bluejeans.com/441850968 > > Gluster-users mailing list > Gluster-users@gluster.org > https://lists.gluster.org/mailman/listinfo/gluster-users > smime.p7s Description: S/MIME cryptographic signature Community Meeting Calendar: APAC Schedule - Every 2nd and 4th Tuesday at 11:30 AM IST Bridge: https://bluejeans.com/441850968 NA/EMEA Schedule - Every 1st and 3rd Tuesday at 01:00 PM EDT Bridge: https://bluejeans.com/441850968 Gluster-users mailing list Gluster-users@gluster.org https://lists.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] Trying to fix files that don't want to heal
- Original Message - From: "Gudrun Mareike Amedick" To: "Ashish Pandey" Cc: "Gluster-users" Sent: Friday, November 29, 2019 8:45:13 PM Subject: Re: [Gluster-users] Trying to fix files that don't want to heal Hi Ashish, thanks for your reply. To fulfill the "no IO"-requirement, I'll have to wait until second week of December (9th – 14th). We originally planned to update GlusterFS from 4.1.7 to 5 and then to 6 in December. Should we do that upgrade before or after running those scripts? The best >> It will be best if you could do it before upgrading to newer version. BTW, why are you not planing to upgrade to gluster 7? Kind regards GudrunAm Freitag, den 29.11.2019, 00:38 -0500 schrieb Ashish Pandey: > Hey Gudrun, > > Could you please try to use the scripts and try to resolve it. > We have written some scripts and it is in final phase to get merge - > https://review.gluster.org/#/c/glusterfs/+/23380/ > > You can find the steps to use these scripts in README.md file > > --- > Ashish > > From: "Gudrun Mareike Amedick" > To: "Gluster-users" > Sent: Thursday, November 28, 2019 3:57:18 PM > Subject: [Gluster-users] Trying to fix files that don't want to heal > > Hi, > > I have a distributed dispersed volume with files that don't want to heal. I'm > trying to fix them manually. > > I'm currently working on a file that is present on all bricks, GFID exists in > .glusterfs-structure and getfattr shows identical attributes for all > files. They look like this: > > # getfattr -m. -d -e hex $brick/somepath/libssl.so.1.1 > getfattr: Removing leading '/' from absolute path names > # file: $brick/$somepath/libssl.so.1.1 > trusted.ec.config=0x080602000200 > trusted.ec.dirty=0x0001 > trusted.ec.size=0x000a > trusted.ec.version=0x00040005 > trusted.gfid=0xdd7dd64f6bb34b5f891a5e32fe83874f > trusted.gfid2path.0c3a5b76c518ef60=0x34663064396234332d343730342d343634352d613834342d3338303532336137346632662f6c696273736c2e736f2e312e31 > > trusted.gfid2path.578ce2ec37aa0f9d=0x31636136323433342d396132642d343039362d616265352d6463353065613131333066632f6c696273736c2e736f2e312e31 > > trusted.glusterfs.quota.1ca62434-9a2d-4096-abe5-dc50ea1130fc.contri.3=0x00029201 > > trusted.glusterfs.quota.4f0d9b43-4704-4645-a844-380523a74f2f.contri.3=0x00029201 > > trusted.pgfid.1ca62434-9a2d-4096-abe5-dc50ea1130fc=0x0001 > trusted.pgfid.4f0d9b43-4704-4645-a844-380523a74f2f=0x0001 > > pgfid is "parent gfid" right? Both GFID's refer to a dir in my volume, both > of those dirs contain a file named libssl.so.1.1. They seem to be > hardlinks: > > find $brick/$somepath -samefile $brick/$someotherpath/libssl.so.1.1 > $brick/$somepath/libssl.so.1 > > This exceeds the limits of my GlusterFS knowledge. Is that something that > can/should happen? If not, is it the reason that file won't heal and how > do > I fix that? > > Kind regards > > Gudrun Amedick > > > Community Meeting Calendar: > > APAC Schedule - > Every 2nd and 4th Tuesday at 11:30 AM IST > Bridge: https://bluejeans.com/441850968 > > NA/EMEA Schedule - > Every 1st and 3rd Tuesday at 01:00 PM EDT > Bridge: https://bluejeans.com/441850968 > > Gluster-users mailing list > Gluster-users@gluster.org > https://lists.gluster.org/mailman/listinfo/gluster-users > Community Meeting Calendar: APAC Schedule - Every 2nd and 4th Tuesday at 11:30 AM IST Bridge: https://bluejeans.com/441850968 NA/EMEA Schedule - Every 1st and 3rd Tuesday at 01:00 PM EDT Bridge: https://bluejeans.com/441850968 Gluster-users mailing list Gluster-users@gluster.org https://lists.gluster.org/mailman/listinfo/gluster-users Community Meeting Calendar: APAC Schedule - Every 2nd and 4th Tuesday at 11:30 AM IST Bridge: https://bluejeans.com/441850968 NA/EMEA Schedule - Every 1st and 3rd Tuesday at 01:00 PM EDT Bridge: https://bluejeans.com/441850968 Gluster-users mailing list Gluster-users@gluster.org https://lists.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] Trying to fix files that don't want to heal
Am Montag, den 02.12.2019, 02:15 -0500 schrieb Ashish Pandey: > > > From: "Gudrun Mareike Amedick" > To: "Ashish Pandey" > Cc: "Gluster-users" > Sent: Friday, November 29, 2019 8:45:13 PM > Subject: Re: [Gluster-users] Trying to fix files that don't want to heal > > Hi Ashish, > > thanks for your reply. To fulfill the "no IO"-requirement, I'll have to wait > until second week of December (9th – 14th). > > We originally planned to update GlusterFS from 4.1.7 to 5 and then to 6 in > December. Should we do that upgrade before or after running those > scripts? > The best > > >> It will be best if you could do it before upgrading to newer version. > BTW, why are you not planing to upgrade to gluster 7? Okay, will do. We have ~400TB-volume. Restoring it is no fun. As a result, we have a pretty conservative updating schedule. > > > Kind regards > > GudrunAm Freitag, den 29.11.2019, 00:38 -0500 schrieb Ashish Pandey: > > Hey Gudrun, > > > > Could you please try to use the scripts and try to resolve it. > > We have written some scripts and it is in final phase to get merge - > > https://review.gluster.org/#/c/glusterfs/+/23380/ > > > > You can find the steps to use these scripts in README.md file > > > > --- > > Ashish > > > > From: "Gudrun Mareike Amedick" > > To: "Gluster-users" > > Sent: Thursday, November 28, 2019 3:57:18 PM > > Subject: [Gluster-users] Trying to fix files that don't want to heal > > > > Hi, > > > > I have a distributed dispersed volume with files that don't want to heal. > > I'm trying to fix them manually. > > > > I'm currently working on a file that is present on all bricks, GFID exists > > in .glusterfs-structure and getfattr shows identical attributes for all > > files. They look like this: > > > > # getfattr -m. -d -e hex $brick/somepath/libssl.so.1.1 > > getfattr: Removing leading '/' from absolute path names > > # file: $brick/$somepath/libssl.so.1.1 > > trusted.ec.config=0x080602000200 > > trusted.ec.dirty=0x0001 > > trusted.ec.size=0x000a > > trusted.ec.version=0x00040005 > > trusted.gfid=0xdd7dd64f6bb34b5f891a5e32fe83874f > > trusted.gfid2path.0c3a5b76c518ef60=0x34663064396234332d343730342d343634352d613834342d3338303532336137346632662f6c696273736c2e736f2e312e31 > > trusted.gfid2path.578ce2ec37aa0f9d=0x31636136323433342d396132642d343039362d616265352d6463353065613131333066632f6c696273736c2e736f2e312e31 > > trusted.glusterfs.quota.1ca62434-9a2d-4096-abe5-dc50ea1130fc.contri.3=0x00029201 > > trusted.glusterfs.quota.4f0d9b43-4704-4645-a844-380523a74f2f.contri.3=0x00029201 > > trusted.pgfid.1ca62434-9a2d-4096-abe5-dc50ea1130fc=0x0001 > > trusted.pgfid.4f0d9b43-4704-4645-a844-380523a74f2f=0x0001 > > > > pgfid is "parent gfid" right? Both GFID's refer to a dir in my volume, both > > of those dirs contain a file named libssl.so.1.1. They seem to be > > hardlinks: > > > > find $brick/$somepath -samefile $brick/$someotherpath/libssl.so.1.1 > > $brick/$somepath/libssl.so.1 > > > > This exceeds the limits of my GlusterFS knowledge. Is that something that > > can/should happen? If not, is it the reason that file won't heal and how > > do > > I fix that? > > > > Kind regards > > > > Gudrun Amedick > > > > > > Community Meeting Calendar: > > > > APAC Schedule - > > Every 2nd and 4th Tuesday at 11:30 AM IST > > Bridge: https://bluejeans.com/441850968 > > > > NA/EMEA Schedule - > > Every 1st and 3rd Tuesday at 01:00 PM EDT > > Bridge: https://bluejeans.com/441850968 > > > > Gluster-users mailing list > > Gluster-users@gluster.org > > https://lists.gluster.org/mailman/listinfo/gluster-users > > > > > Community Meeting Calendar: > > APAC Schedule - > Every 2nd and 4th Tuesday at 11:30 AM IST > Bridge: https://bluejeans.com/441850968 > > NA/EMEA Schedule - > Every 1st and 3rd Tuesday at 01:00 PM EDT > Bridge: https://bluejeans.com/441850968 > > Gluster-users mailing list > Gluster-users@gluster.org > https://lists.gluster.org/mailman/listinfo/gluster-users > smime.p7s Description: S/MIME cryptographic signature Community Meeting Calendar: APAC Schedule - Every 2nd and 4th Tuesday at 11:30 AM IST Bridge: https://bluejeans.com/441850968 NA/EMEA Schedule - Every 1st and 3rd Tuesday at 01:00 PM EDT Bridge: https://bluejeans.com/441850968 Gluster-users mailing list Gluster-users@gluster.org https://lists.gluster.org/mailman/listinfo/gluster-users
Re: [Gluster-users] Trying to fix files that don't want to heal
Hi, so I ran gfid_needing_heal_parallel_new.sh. First try, I got a whole bunch of "ssh_exchange_identification: Connection closed by remote host", increasing MaxStartups in sshd_config fixed that. The results were the same in both runs though, so that might be a cosmetic problem. I ended up with a file that contains lines like this for each broken file $FILE /$SERVER and $BRICK iterate over all servers and bricks respectively): file:$NUMBER|$NUMBER|$SERVER:/$BRICK#$FILE, So I'm feeding that into correct_pending_heals.sh and I should be done, right? I have an entry in my heal info for a file that doesn't exist anymore, I deleted it a while ago. Looking into the dir structure revealed a stray link file (permissions -T) for that file on some servers. Not sure why that wasn't deleted. It's in potential_heal though, the line starts with "empty|$SERVER(..)". I'll delete that line from potential_heal. It might be an interesting edge case though. Kind regards Gudrun Am Montag, den 02.12.2019, 02:15 -0500 schrieb Ashish Pandey: > > > From: "Gudrun Mareike Amedick" > To: "Ashish Pandey" > Cc: "Gluster-users" > Sent: Friday, November 29, 2019 8:45:13 PM > Subject: Re: [Gluster-users] Trying to fix files that don't want to heal > > Hi Ashish, > > thanks for your reply. To fulfill the "no IO"-requirement, I'll have to wait > until second week of December (9th – 14th). > > We originally planned to update GlusterFS from 4.1.7 to 5 and then to 6 in > December. Should we do that upgrade before or after running those > scripts? > The best > > >> It will be best if you could do it before upgrading to newer version. > BTW, why are you not planing to upgrade to gluster 7? > > > Kind regards > > GudrunAm Freitag, den 29.11.2019, 00:38 -0500 schrieb Ashish Pandey: > > Hey Gudrun, > > > > Could you please try to use the scripts and try to resolve it. > > We have written some scripts and it is in final phase to get merge - > > https://review.gluster.org/#/c/glusterfs/+/23380/ > > > > You can find the steps to use these scripts in README.md file > > > > --- > > Ashish > > > > From: "Gudrun Mareike Amedick" > > To: "Gluster-users" > > Sent: Thursday, November 28, 2019 3:57:18 PM > > Subject: [Gluster-users] Trying to fix files that don't want to heal > > > > Hi, > > > > I have a distributed dispersed volume with files that don't want to heal. > > I'm trying to fix them manually. > > > > I'm currently working on a file that is present on all bricks, GFID exists > > in .glusterfs-structure and getfattr shows identical attributes for all > > files. They look like this: > > > > # getfattr -m. -d -e hex $brick/somepath/libssl.so.1.1 > > getfattr: Removing leading '/' from absolute path names > > # file: $brick/$somepath/libssl.so.1.1 > > trusted.ec.config=0x080602000200 > > trusted.ec.dirty=0x0001 > > trusted.ec.size=0x000a > > trusted.ec.version=0x00040005 > > trusted.gfid=0xdd7dd64f6bb34b5f891a5e32fe83874f > > trusted.gfid2path.0c3a5b76c518ef60=0x34663064396234332d343730342d343634352d613834342d3338303532336137346632662f6c696273736c2e736f2e312e31 > > trusted.gfid2path.578ce2ec37aa0f9d=0x31636136323433342d396132642d343039362d616265352d6463353065613131333066632f6c696273736c2e736f2e312e31 > > trusted.glusterfs.quota.1ca62434-9a2d-4096-abe5-dc50ea1130fc.contri.3=0x00029201 > > trusted.glusterfs.quota.4f0d9b43-4704-4645-a844-380523a74f2f.contri.3=0x00029201 > > trusted.pgfid.1ca62434-9a2d-4096-abe5-dc50ea1130fc=0x0001 > > trusted.pgfid.4f0d9b43-4704-4645-a844-380523a74f2f=0x0001 > > > > pgfid is "parent gfid" right? Both GFID's refer to a dir in my volume, both > > of those dirs contain a file named libssl.so.1.1. They seem to be > > hardlinks: > > > > find $brick/$somepath -samefile $brick/$someotherpath/libssl.so.1.1 > > $brick/$somepath/libssl.so.1 > > > > This exceeds the limits of my GlusterFS knowledge. Is that something that > > can/should happen? If not, is it the reason that file won't heal and how > > do > > I fix that? > > > > Kind regards > > > > Gudrun Amedick > > > > > > Community Meeting Calendar: > > > > APAC Schedule - > > Every 2nd and 4th Tuesday at 11:30 AM IST > > Bridge: https://bluejeans.com/441850968 > >