The root cause of this error is that the pool of reserved GDT blocks is not available, or the filesystem does not support online resizing. Note that the Ext3 and Ext4 filesystem metadata layout is fixed. mkfs reserves space for future disk resizing, but that is only 1024x the filesystem size during initial disk creation or the upper block count limit of 2^32, priority given to the lowest. The third root cause of the error is that the journal is too small.
- To resolve this error:
check if online resizing is available for the filesystem. You can check this with resize_inode in the dumpe2fs output. If resize_inode text is not present in the dumpe2fs output given below, then the filesystem does not support online resizing. It would be best if you then unmounted the filesystem and then resize it.
dumpe2fs /dev/vg_test/lv_ext3 | grep -i features
Filesystem features: has_journal ext_attr resize_inode dir_index filetype needs_recovery extent flex_bg sparse_super large_file huge_file uninit_bg dir_nlink extra_isize
Journal features: journal_incompat_revoke
- Check the output of dumpe2fs. If the Reserved GDT block line is missing, then the GDT block count is 0. The solution, in this case, is to resize the filesystem offline, i.e., after unmounting the filesystem.
- The third scenario is when resize fails in both offline and online mode. The solution, in this case, is to remove the journal while the filesystem is offline and recreate the journal with a larger size.
First, check the size of the journal with;
dumpe2fs /dev/vg_test/lv_ext3 | grep Journal\ size
Journal size: 32M
Now remove the journal with
tune2fs -O ^has_journal /dev/vg_test/lv_ext3
Creating journal inode: done
Verify the journal size you just create with;
dumpe2fs /dev/vg_test/lv_ext3 | grep Journal\ size
Journal size: 148M
Now resize with resize2fs /dev/vg_test/lv_ext3
If you still get errors, use the option -J size=journal-size, where journal-size is in megabytes.
