Remounting an ACFS filesystem

Another story that kept me awake like this one will be the todays topic. By some funny coincidence it’s also about the same customer. This time I was called because a very important ACFS filesystem could not be mounted. The customer tried to change the group- and user-ID of the Grid Infrastructure Home which was not a good idea at all. So they reverted all the changes but an ACFS filesystem did not come up again. That was the moment they called me.

So I first checked the ASM volumes. There was only one volume in “disabled” state.

[oracle@server02 ~]$ asmcmd volinfo --all
Diskgroup Name: DATA

         Volume Name: U27
         Volume Device: /dev/asm/u27-510
         State: DISABLED
         Size (MB): 31815680
         Resize Unit (MB): 64
         Redundancy: UNPROT
         Stripe Columns: 8
         Stripe Width (K): 1024
         Usage: ACFS
         Mountpath: /u27

The obvious idea was to re-enable to volume. But that didn’t succeed.

[oracle@server02 ~]$ asmcmd volenable --all
ORA-15032: not all alterations performed
ORA-15489: conflicting volume already open on this node (DBD ERROR: OCIStmtExecute)

[oracle@server02 ~]$ oerr ora 15489
15489, 00000, "conflicting volume already open on this node"
// *Cause: An ASM volume was not closed prior to instance shutdown
// on node specified. It is conflicting with the volume that
// is to be enabled.
// *Action: Close the ASM volume either via unmount or node restart and
// retry the operation.
//

Ok, we have a disabled volume which is still in use somehow. It really is, see this:

[root@server02 ~]# mount|grep u27
/dev/asm/u27-510 on /u27 type acfs (rw,relatime,device,ordered)
[root@server02 ~]# ll /u27
ls: cannot open directory /u27: Input/output error
[root@server02 ~]# umount /u27
umount: /u27: target is busy.
        (In some cases useful info about processes that use
         the device is found by lsof(8) or fuser(1))
umount.acfs: CLSU-00107: operating system function: OfsWaitProc; failed with error data: 32; at location: OWPR_1
umount.acfs: CLSU-00101: operating system error message: Broken pipe
umount.acfs: ACFS-04151: unmount of mount point /u27 failed

I can’t use the filesystem and cannot unmount the filesystem either. So what is preventing me from unmounting the filesytem?

[root@server02 ~]# fuser /u27
[root@server02 ~]# lsof |grep  /u27
[root@server02 ~]# umount /u27
umount: /u27: target is busy.
        (In some cases useful info about processes that use
         the device is found by lsof(8) or fuser(1))
umount.acfs: CLSU-00107: operating system function: OfsWaitProc; failed with error data: 32; at location: OWPR_1
umount.acfs: CLSU-00101: operating system error message: Broken pipe
umount.acfs: ACFS-04151: unmount of mount point /u27 failed
[root@server02 ~]# lsof |grep u27-510

At this point the customer and I agreed to simply reboot the system. Beside the fact, that the system did not come up in the first place due to some misconfiguration in the grub.conf, this finally solved the issue.

I just needed to load the kernel modules since this was a stand-alone installation, enable the volume and finally mount the filesysem again.

[root@server02 ~]# lsmod|grep advm
[root@server02 ~]# $GRID_HOME/bin/acfsload start
[root@server02 ~]# su - oracle
[oracle@server02 ~]$ asmcmd  volenable --all
[root@server02 ~]# /sbin/fsck -a -y -t acfs /dev/asm/u27-510
[root@server02 ~]# mount -t acfs /dev/asm/u27-510 /u27

Customer happy, their Golden Gate Trails where back again.

Auto Resize and Shrink ACFS Volumes

Today I want to write a little about ACFS. In case you are using the latest and greatest Oracle Grid Infrastructure and you still run 11.2 databases, you are required to use ACFS. This especially applies to the Database Appliances, ExaDatas and for the Cloud Services, if you use Grid Infrastructure there. Having that, you might run into situations where your ACFS filesystem grows which it does automatically. But what if you want to shrink the filesystem to it’s original or at least a smaller size? With version 18.3 this should work quite well. Version before did had some issues when there was no contigous free space at the end of the volume to do the shrinking. Read more about that in MOS Note “ODA (Oracle Database Appliance): Resizing ACFS DBStorage File System (Doc ID 1990430.1)“.
Basically, there are two major topics. First, the shrinking of ACFS volumes and second, limiting the auto resize of ACFS.

Current situation

Given a customers ODA, this is the current situation.

[root@oda-x7m ~]# df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/mapper/VolGroupSys-LogVolRoot
                       30G  3.4G   25G  13% /
tmpfs                 189G  1.3G  187G   1% /dev/shm
/dev/md0              477M   40M  412M   9% /boot
/dev/sda1             500M  320K  500M   1% /boot/efi
/dev/mapper/VolGroupSys-LogVolOpt
                       59G   24G   33G  43% /opt
/dev/mapper/VolGroupSys-LogVolU01
                       99G   78G   16G  84% /u01
/dev/asm/commonstore-241
                      5.0G  319M  4.7G   7% /opt/oracle/dcs/commonstore
/dev/asm/dattfstar01-241
                      170G  158G   13G  93% /u02/app/oracle/oradata/TFSTAR01
/dev/asm/reco-135     883G  693G  191G  79% /u03/app/oracle
/dev/asm/datxvgvg-241
                      100G  5.9G   95G   6% /u02/app/oracle/oradata/XVGVG
/dev/asm/datxdovg-241
                      410G  399G   12G  98% /u02/app/oracle/oradata/XDOVG
/dev/asm/datxvgovg-241
                      100G  5.7G   95G   6% /u02/app/oracle/oradata/XVGOVG
/dev/asm/datxdoovg-241
                      390G  376G   15G  97% /u02/app/oracle/oradata/XDOOVG
/dev/asm/datxvgfg-241
                      100G  5.1G   95G   6% /u02/app/oracle/oradata/XVGFG
/dev/asm/datxdofg-241
                      100G   46G   55G  46% /u02/app/oracle/oradata/XDOFG
/dev/asm/datsfstar01-241
                      100G  6.3G   94G   7% /u02/app/oracle/oradata/SFSTAR01
/dev/asm/dattvgfg01-241
                      100G  5.1G   95G   6% /u02/app/oracle/oradata/TVGFG01
/dev/asm/dattfstar03-241
                      100G  4.9G   96G   5% /u02/app/oracle/oradata/TFSTAR03
/dev/asm/dattfstar04-241
                      100G  7.0G   94G   7% /u02/app/oracle/oradata/TFSTAR04
/dev/asm/dattfstar02-241
                      1.9T  1.9T   76M 100% /u02/app/oracle/oradata/TFSTAR02
/dev/asm/dattdomea01-241
                      100G   20G   81G  20% /u02/app/oracle/oradata/TDOMEA01
/dev/asm/dattfstar05-241
                      100G  5.2G   95G   6% /u02/app/oracle/oradata/TFSTAR05
/dev/asm/datifstar01-241
                      100G  5.1G   95G   6% /u02/app/oracle/oradata/IFSTAR01A
/dev/asm/datrfstar01-241
                      1.1T  1.1T  9.4G 100% /u02/app/oracle/oradata/RFSTAR01A
jus-srv-rz50.justiz.sachsen.de:/u01/app/DB-dumps
                      3.0T  2.0T  924G  68% /u99/nfs/DB-dumps
/dev/asm/datsfstar02-241
                      100G  3.6G   97G   4% /u02/app/oracle/oradata/SFSTAR02

Shrinking an ACFS Filesystem

As to be seen above, there are quite a lot of databases that have a lot of free space in their filesystems. The initial size is defined by the shapes that an ODA provides. But in our case, even 100GB are too much, so we wanted to reduce the occupied space.
First, we get some basic information.

[oracle@oda-x7m ~]$ /sbin/advmutil volinfo /dev/asm/datxvgvg-241
Device: /dev/asm/datxvgvg-241
Size (MB): 102400
Resize Increment (MB): 512
Redundancy: mirror
Stripe Columns: 8
Stripe Width (KB): 1024
Disk Group: DATA
Volume: DATXVGVG
Compatible.advm: 12.2.0.1.0

The “resize” feature requires the ADVM compatibility to be at least 12.2.0.1, so this is the case. Now let’s do the shrink.

[oracle@oda-x7m ~]$ acfsutil size -h
Usage: acfsutil [-h] size [[-|+]nnn[K|M|G|T|P]] [-a nnn[K|M|G|T|P]] [-x nnn[K|M|G|T|P]] [-d ]  
                                         - Resize file system and configure auto-resize
                [-a]                     - Auto-resize increment
                [-x]                     - Auto-resize maximum
                [-d]                     - Resize only this device

[oracle@oda-x7m ~]$ acfsutil size -80G /u02/app/oracle/oradata/XVGVG
acfsutil size: new file system size: 21474836480 (20480MB)

After that, the volume is at the desired size of 20GB.

[oracle@oda-x7m ~]$ /sbin/advmutil volinfo /dev/asm/datxvgvg-241
Device: /dev/asm/datxvgvg-241
Size (MB): 20480
Resize Increment (MB): 512
Redundancy: mirror
Stripe Columns: 8
Stripe Width (KB): 1024
Disk Group: DATA
Volume: DATXVGVG
Compatible.advm: 12.2.0.1.0

Limiting auto increment of ACFS

The automatic increment of ACFS volumes has two implications. First, it makes monitoring difficult because you never know is the filesystem is really full or will it automatically resize if needed. To figure out, if you are having a critical issue, you need to perform several steps. Implementing this with a monitoring system will be cumbersome in most cases. The other thing is, managing the storage resources is not straight-forward. You might overcommit your storage. My approach and advise is, to disable automatic resizing by setting the ACFS size and maximum size to the value, that you commited for this specific database. Let’s check the setup before doing that.

[oracle@oda-x7m ~]$ acfsutil info fs /u02/app/oracle/oradata/ICM01A
/u02/app/oracle/oradata/ICM01A
    ACFS Version: 12.2.0.1.0
    on-disk version:       46.0
    compatible.advm:       12.2.0.0.0
    ACFS compatibility:    12.2.0.0.0
    flags:        MountPoint,Available,AutoResizeEnabled
    mount time:   Fri Mar 15 10:09:28 2019
    mount sequence number: 4
    allocation unit:       4096
    metadata block size:   4096
    volumes:      1
    total size:   107374182400  ( 100.00 GB )
    total free:   91318923264  (  85.05 GB )
    auto-resize increment:   10737418240  (  10.00 GB )
    auto-resize maximum:     0  ( 0.00 )
    file entry table allocation: 262144
    primary volume: /dev/asm/daticm01-411
        label:
        state:                 Available
        major, minor:          246, 210435
        logical sector size:   512
        size:                  107374182400  ( 100.00 GB )
        free:                  91318923264  (  85.05 GB )
        metadata read I/O count:         54838
        metadata write I/O count:        51022
        total metadata bytes read:       242651136  ( 231.41 MB )
        total metadata bytes written:    233385984  ( 222.57 MB )
        ADVM diskgroup:        DATA
        ADVM resize increment: 536870912
        ADVM redundancy:       mirror
        ADVM stripe columns:   8
        ADVM stripe width:     1048576
    number of snapshots:  0
    snapshot space usage: 0  ( 0.00 )
    replication status: DISABLED
    compression status: DISABLED

So I set the maximum size to the current size.

[oracle@oda-x7m ~]$ acfsutil size -x 100G /u02/app/oracle/oradata/ICM01A
acfsutil size: ACFS-03642: successfully updated auto-resize settings

Just crosscheck that.

[oracle@oda-x7m ~]$ acfsutil info fs /u02/app/oracle/oradata/ICM01A
/u02/app/oracle/oradata/ICM01A
    ACFS Version: 12.2.0.1.0
    on-disk version:       46.0
    compatible.advm:       12.2.0.0.0
    ACFS compatibility:    12.2.0.0.0
    flags:        MountPoint,Available,AutoResizeEnabled
    mount time:   Fri Mar 15 10:09:28 2019
    mount sequence number: 4
    allocation unit:       4096
    metadata block size:   4096
    volumes:      1
    total size:   107374182400  ( 100.00 GB )
    total free:   91318923264  (  85.05 GB )
    auto-resize increment:   10737418240  (  10.00 GB )
    auto-resize maximum:     107374182400  ( 100.00 GB )
    file entry table allocation: 262144
    primary volume: /dev/asm/daticm01-411
        label:
        state:                 Available
        major, minor:          246, 210435
        logical sector size:   512
        size:                  107374182400  ( 100.00 GB )
        free:                  91318923264  (  85.05 GB )
        metadata read I/O count:         54840
        metadata write I/O count:        51022
        total metadata bytes read:       242659328  ( 231.42 MB )
        total metadata bytes written:    233385984  ( 222.57 MB )
        ADVM diskgroup:        DATA
        ADVM resize increment: 536870912
        ADVM redundancy:       mirror
        ADVM stripe columns:   8
        ADVM stripe width:     1048576
    number of snapshots:  0
    snapshot space usage: 0  ( 0.00 )
    replication status: DISABLED
    compression status: DISABLED

You could also combine the resize operation which returns the final ACFS size which in turn can be used for setting the maximum size.
And that’s basically it, there is nothing more to do.