mdadm/md-check how to check array integrity without rebuilding

This doesn't seem to be widely known (maybe it's in some documentation that none of us read though) but there's an easy way to check the integrity of any mdadm array:

sudo echo check > /sys/block/md0/md/sync_action
-bash: /sys/block/md0/md/sync_action: Permission denied

sudo will never work, this only works as root since echo is not actually a binary/command.  It is built-into bash.


/sys/devices/virtual/block/md0/md# echo check >sync_action
 cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md_d1 : inactive sdc2[1](S)
      20490816 blocks
      
md0 : active raid1 sda1[0] sdc1[1]
      943730240 blocks [2/2] [UU]
      [>....................]  check =  0.0% (424768/943730240) finish=111.0min speed=141589K/sec
     
unused devices: <none>

 

In dmesg you'll see the following:

[ 1761.813250] md: data-check of RAID array md0
[ 1761.813257] md: minimum _guaranteed_  speed: 1000 KB/sec/disk.
[ 1761.813263] md: using maximum available idle IO bandwidth (but not more than 200000 KB/sec) for data-check.
[ 1761.813273] md: using 128k window, over a total of 943730240 blocks.
[11451.021752] md: md0: data-check done.
[32917.662549] md: data-check of RAID array md0
[32917.662556] md: minimum _guaranteed_  speed: 1000 KB/sec/disk.
[32917.662562] md: using maximum available idle IO bandwidth (but not more than 200000 KB/sec) for data-check.
[32917.662571] md: using 128k window, over a total of 943730240 blocks.
 

When it's done you'll see this:

[11451.021752] md: md0: data-check done.

 

When there's corruption I have no idea what you'd see, maybe in mdstat you'll see a bad disk is off-line and marked as failed and an entry in dmesg/messages that the check is failed or completed with errors.  Just a guess, I've never seen any example of reference on the net about what happens or what is said if there is an inconsistency found.  Distributions like Ubuntu have a cronjob/script that check the array once a week though.


Tags:

mdadm, md, array, rebuildingthis, doesn, widely, documentation, sudo, echo, sys, sync_action, bash, binary, devices, virtual, proc, mdstat, personalities, linear, multipath, raid, md_d, inactive, sdc, active, sda, uu, min, unused, dmesg, ll, _guaranteed_, kb, disk, maximum, idle, io, bandwidth, corruption, marked, entry, completed, errors, ve, reference, inconsistency, distributions, ubuntu, cronjob,

Latest Articles

  • FreePBX 17 How To Add a Trunk
  • Docker Container Onboot Policy - How to make sure a container is always running
  • FreePBX 17 How To Add Phones / Extensions and Register
  • Warning: The driver descriptor says the physical block size is 2048 bytes, but Linux says it is 512 bytes. solution
  • Cisco How To Use a Third Party SIP Phone (eg. Avaya, 3CX)
  • Cisco Unified Communication Manager (CUCM) - How To Add Phones
  • pptp / pptpd not working in DD-WRT iptables / router
  • systemd-journald high memory usage solution
  • How to Install FreePBX 17 in Linux Debian Ubuntu Mint Guide
  • How To Install Cisco's CUCM (Cisco Unified Communication Manager) 12 Guide
  • Linux Ubuntu Redhat How To Extract Images from PDF
  • Linux and Windows Dual Boot Issue NIC Won't work After Booting Windows
  • Cisco CME How To Enable ACD hunt groups
  • How to install gns3 on Linux Ubuntu Mint
  • How to convert audio for Asterisk .wav format
  • Using Cisco CME Router with Asterisk as a dial-peer
  • Cisco CME How To Configure SIP Trunk VOIP
  • Virtualbox host Only Network Error Failed to save host network interface parameter - Cannot change gateway IP of host only network
  • Cisco CME and C7200 Router Testing and Learning Environment on Ubuntu 20+ Setup Tutorial Guide
  • Abusive IP ranges blacklist