Linux High IO load.. what to check for trouble shooting?

When you look at the CPU activity of your computer, one of the parameters is the iowait. This value shows how much time your CPU wastes while it is waiting for I/O operations for complete. These include disk read/write operations, network, IPC, etc. Is this behavior a problem and, if so, what causes it and how to fix it? One one of the popular Unix-related forums one “genius” wrote:

The iowait “problem” is funny. It’s like when people complain that Linux is “using all my memory”. Yeah, no shit. You should be upset if you are copying files and your computer is /not/ in 100% iowait.

In reality, 100% iowait indicates that there is a problem and in most cases – a big problem that may even lead to data loss. Essentially, there is a bottleneck somewhere in the system. Maybe one of your disks is getting ready to die; or, perhaps, the NIC firmware is having problems with the latest kernel upgrade you installed. The troubleshooting process starts with the potentially more serious possibility: bad disk.

Take a quick look at /etc/messages, /etc/dmesg, /etc/boot.log and any other system log files. You are looking for disk I/O errors, failed read/write operations, bad sectors – anything that indicates a hardware problem with a disk. If you don’t find anything, look for IRQ and disk controller errors. Also look for memory errors and kernel panics. The three most likely culprits of high iowait are: bad disk, faulty memory and network problems.

If you still see nothing relevant, it is time to test your system. If possible, kick all the users off the box, shut down Web server, database and any other user application. Log in via command line and stop XDM.

Open three shell windows: run “top” in one, “iostat -x 1? in the other and “find /etc -type f -print” in the third. Make sure you can see all three windows at the same time. This is a simple test that should generate some I/O activity on the system disk. Repeat this process for other disks. If you see iowait hovering near 100%, chance are you have a problem but we don’t know what it is yet. However, now we do know that network is probably not the cause.

deathstar:/ # iostat -x 1
Linux 2.6.5-7.201-default (deathstar) 12/20/08

avg-cpu: %user %nice %sys %iowait %idle
2.83 0.42 1.45 9.11 86.20

Device: rrqm/s wrqm/s r/s w/s rsec/s wsec/s rkB/s wkB/s avgrq-sz avgqu-sz await svctm %util
hda 40.63 66.34 27.45 6.04 936.50 581.23 468.25 290.61 45.32 2.42 72.16 2.22 7.42
hdc 0.01 0.00 0.01 0.00 0.03 0.00 0.02 0.00 4.02 0.00 1.17 1.17 0.00
sda 0.09 2.32 4.15 1.33 71.56 29.23 35.78 14.62 18.37 0.65 118.49 6.39 3.51
sdb 3.47 0.00 1.90 0.00 15.32 0.01 7.66 0.01 8.08 0.74 391.31 5.68 1.08
fd0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 2.00 0.00 45.00 45.00 0.00

deathstar:/ # top
top – 21:28:28 up 1:22, 2 users, load average: 0.09, 0.14, 0.16
Tasks: 77 total, 1 running, 76 sleeping, 0 stopped, 0 zombie
Cpu(s): 2.8% us, 1.3% sy, 0.4% ni, 86.2% id, 9.1% wa, 0.1% hi, 0.0% si
Mem: 508644k total, 503612k used, 5032k free, 34052k buffers
Swap: 1020088k total, 458980k used, 561108k free, 16012k cached

1 root 16 0 640 56 28 S 0.0 0.0 0:05.14 init
2 root 34 19 0 0 0 S 0.0 0.0 0:00.00 ksoftirqd/0
3 root 5 -10 0 0 0 S 0.0 0.0 0:00.09 events/0
4 root 5 -10 0 0 0 S 0.0 0.0 0:00.00 khelper

Next step, lets stress out your CPU but not the disks. The command below will try to create an endless zip file in /dev/null. This generates no disk activity, but loads the CPU. Continue running “top” and “iostat -x 1? in the other two windows.

cat /dev/zero | bzip2 -c > /dev/null

If you see high CPU load but low iowait, we can eliminate CPU issues, IRQ conflicts, and faulty memory. Just to be on the safe side, let’s test memory anyway:

deathstar:/ # free
total used free shared buffers cached
Mem: 508644 503504 5140 0 37036 48968
-/+ buffers/cache: 417500 91144
Swap: 1020088 516196 503892

This server has 508644Kb of RAM. Use the corresponding value for the following test:

deathstar:/ # dd if=/dev/hda2 bs=508644 of=/backups/memtest count=1050
1050+0 records in
1050+0 records out

deathstar:/ # md5sum /backups/memtest ; md5sum /backups/memtest ; md5sum /backups/memtest
04762ff36b2231aac75754ab9c1a564a /backups/memtest
04762ff36b2231aac75754ab9c1a564a /backups/memtest
04762ff36b2231aac75754ab9c1a564a /backups/memtest

The three MD5 values above should be identical. If they are not – your system has a faulty RAM chip.

When you have eliminated hardware problems as possible causes of high iowait, the next step is to review firmware and drivers. You are particularly interested in disk controller firmware: unstable performance and no error messages are the signs of a firmware problem. Try really hard to remember if you made any system changes recently, especially something that required a reboot – like kernel upgrade, for example. If this is the case, roll back the upgrade or search for upgrade firmware. You should grab a copy of Sysinfo (free 30-day trial) to help you identify makes and models of your disks, controllers, etc.

While your disks and controllers may be tip-top, your may have a problem with a filesystem. Even if you see high iowait when accessing any filesystem, you should still check out the partition where /var is mounted and swap – if there is a problem, it will manifest itself regardless of what your system is doing. But here you will run into a little problem: fsck will not scan a mounted partition and you cannot unmount /var. Let’s say these are your partitions:

deathstar:/ # more /etc/fstab
/dev/hda2 / reiserfs acl,user_xattr 1 1
/dev/hda1 swap swap pri=42 0 0

You need to fsck /dev/hda2 because this is where your /var is mounted. Download KNOPPIX or Ubuntu LiveCD, boot from CD (without installing) and “fsck /dev/hda2? from there. If everything looks clean, shut down your system, take the CD out and boot normally. The next step is to check out swap. If you just run fsck on the swap partition, it will fail:

deathstar:/ # fsck /dev/hda1
fsck 1.34 (25-Jul-2003)
fsck: fsck.swap: not found
fsck: Error 2 while executing fsck.swap for /dev/hda1

You need to disable swap on /dev/hda1 before you can scan it. Before you can do this, you need to add another swap area: you cannot run without any swap space. So, to add swap on the fly, create a swap file (1Gb in this example):

deathstar:/ # dd if=/dev/zero of=/swapfile bs=1024 count=1048576
1048576+0 records in
1048576+0 records out

deathstar:/ # chmod 600 /swapfile

deathstar:/ # ls -lash /swapfile
1.1G -rw——- 1 root root 1.0G Dec 20 22:48 /swapfile

Now you can set up and activate the new swap file:

deathstar:/ # mkswap /swapfile
Setting up swapspace version 1, size = 1073737 kB
deathstar:/ # free
total used free shared buffers cached
Mem: 508644 500996 7648 0 38912 147332
-/+ buffers/cache: 314752 193892
Swap: 1020088 521784 498304
deathstar:/ # swapon /swapfile
deathstar:/ # free
total used free shared buffers cached
Mem: 508644 502232 6412 0 39400 147392
-/+ buffers/cache: 315440 193204
Swap: 2068656 521784 1546872

Now we need to deactivate the original swap partition. This operation may take a couple minutes to complete:

deathstar:/ # swapoff /dev/hda1
deathstar:/ # free
total used free shared buffers cached
Mem: 508644 501624 7020 0 31712 10416
-/+ buffers/cache: 459496 49148
Swap: 1048568 167032 881536

The next step is to create a standard filesystem on the old swap partition, so that fsck has something to scan:

deathstar:/ # mke2fs -c /dev/hda1
mke2fs 1.34 (25-Jul-2003)
Filesystem label=
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
127744 inodes, 255024 blocks
12751 blocks (5.00%) reserved for the super user
First data block=0
8 block groups
32768 blocks per group, 32768 fragments per group
15968 inodes per group
Superblock backups stored on blocks:
32768, 98304, 163840, 229376
Checking for bad blocks (read-only test): done
Writing inode tables: done
Writing superblocks and filesystem accounting information: done

The previous operation already ran fsck and so, if you see no errors, you can now re-activate your original swap space and remove the temporary swap you created:

deathstar:/ # mkswap /dev/hda1
Setting up swapspace version 1, size = 1044574 kB
deathstar:/ # swapon /dev/hda1
deathstar:/ # swapoff /swapfile
deathstar:/ # rm /swapfile
deathstar:/ # free
total used free shared buffers cached
Mem: 508644 503172 5472 0 33668 9256
-/+ buffers/cache: 460248 48396
Swap: 1020088 156300 863788

Anothe command commonly used for analyzing system bottlenecks is vmstat. The following example runs vmstat five times at 2-second intervals:

deathstar:~ # vmstat -S M 2 5
procs ———–memory———- —swap– —–io—- –system– —-cpu—-
r b swpd free buff cache si so bi bo in cs us sy id wa
0 0 15 174 70 58 0 0 189 50 5 6 1 3 94 1
0 0 15 174 70 58 0 0 0 0 1005 35 4 0 96 0
0 1 15 174 70 58 0 0 0 258 1515 45 0 6 88 7
0 0 15 173 71 58 0 0 0 194 1083 24 0 1 83 16
0 0 15 173 71 58 0 0 0 0 1003 19 0 0 100 0

Explanation of vmstat columns:

(a) procs is the process-related fields are:

* r: The number of processes waiting for run time.
* b: The number of processes in uninterruptible sleep.

(b) memory is the memory-related fields are:

* swpd: the amount of virtual memory used.
* free: the amount of idle memory.
* buff: the amount of memory used as buffers.
* cache: the amount of memory used as cache.

(c) swap is swap-related fields are:

* si: Amount of memory swapped in from disk (/s).
* so: Amount of memory swapped to disk (/s).

(d) io is the I/O-related fields are:

* bi: Blocks received from a block device (blocks/s).
* bo: Blocks sent to a block device (blocks/s).

(e) system is the system-related fields are:

* in: The number of interrupts per second, including the clock.
* cs: The number of context switches per second.

(f) cpu is the CPU-related fields are:

These are percentages of total CPU time.

* us: Time spent running non-kernel code. (user time, including nice time)
* sy: Time spent running kernel code. (system time)
* id: Time spent idle. Prior to Linux 2.5.41, this includes IO-wait time.
* wa: Time spent waiting for IO. Prior to Linux 2.5.41, shown as zero.

If you failed to identify the cause of the iowait problem, you should consider the possibility that there is no problem: perhaps your system is handling extra load and running short on resources. Take a look at the running processes and see what’s eating up memory. Perhaps you upgraded an application and now it is using more RAM, which leads to high swapping, which leads to high disk activity, which leads to high iowait.

The solutions are simple:

1. Install more RAM
2. Move swap to another disk or – even better – move it to another disk on a separate controller.
3. Move user applications to another disk/controller and specify default log locations outside of the system disk.

– Jayesh

../../../libraries/libldap/error.c:273: ldap_parse_result: Assertion `r != ((void *)0)’ failed

If you are getting error as mentioned below while doing some operation your linux server bash shell.
../../../libraries/libldap/error.c:273: ldap_parse_result: Assertion `r != ((void *)0)’ failed

Then its due to nss-ldap software running on your server. One of the reason I found and fixed with was nscd service was down on my server restarting it fixed the issue.

Error I saw in logs were..


Oct 28 03:01:27 HOSTNAME nscd: nss_ldap: reconnected to LDAP server ldap:// after 1 attempt
Nov 10 02:49:58 HOSTNAME nscd: nss_ldap: reconnecting to LDAP server (sleeping 4 seconds)…
Nov 10 02:50:14 HOSTNAME nscd: nss_ldap: reconnected to LDAP server ldap:// after 2 attempts
Jan 18 07:45:09 HOSTNAME kernel: nscd[5114]: segfault at 00002b1c735dee78 rip 00002b1b6d4fe885 rsp 000000004185c6d0 error 4

Fix :
[root@HOSTNAME webdocs]# /etc/init.d/nscd status
nscd dead but subsys locked
You have new mail in /var/spool/mail/root
[root@HOSTNAME webdocs]# /etc/init.d/nscd restart
Stopping nscd: [FAILED]
Starting nscd: [ OK ]
[root@HOSTNAME webdocs]# /etc/init.d/nscd status
nscd (pid 30292) is running…
You have new mail in /var/spool/mail/root
[root@HOSTNAME webdocs]#