Sponsored Content
Full Discussion: Iostat inconsistent avgrq-sz
Top Forums UNIX for Advanced & Expert Users Iostat inconsistent avgrq-sz Post 302922726 by maverick_here on Tuesday 28th of October 2014 02:53:02 AM
Old 10-28-2014
RedHat Iostat inconsistent avgrq-sz

Hi Experts,

I have an issue with performance of oracle, the db sits on EBS volumes on and aws EC2 instance. I have captured the iostat data and feed it to iostat_plotter to forward it to the DB administrators. However what I'm trying to understand is the variable avgrq-sz, it is quite in consistent.

The system is running RHEL 6.4 64 bit, Oracle 11g. The block size is 4K .Following is the extract for two devices which are of interest

Code:
Device:         rrqm/s   wrqm/s     r/s     w/s   rsec/s   wsec/s avgrq-sz avgqu-sz   await  svctm  %util
xvdj              0.00    80.55   15.18   59.58  1155.01  1121.04    30.45     1.10   14.74   0.70   5.20
xvdj              0.00     0.00    3.00    2.00    48.00    16.00    12.80     0.03    5.80   5.60   2.80
xvdj              0.00     0.00   14.00  162.00   704.00  1296.00    11.36     0.34    1.94   0.87  15.30
xvdj              0.00     0.00    8.00    4.00   304.00    32.00    28.00     0.15   12.25   5.50   6.60
xvdj              0.00     0.00    0.00    1.00     0.00     8.00     8.00     0.00    1.00   1.00   0.10
xvdj              0.00   114.00    2.00  147.00    96.00  2088.00    14.66     0.86    5.77   0.58   8.60
xvdj              0.00     0.00    0.00   33.00     0.00   264.00     8.00     0.05    1.42   0.48   1.60
xvdj              0.00     0.00    4.00    6.00    64.00    48.00    11.20     0.08    7.60   7.00   7.00
xvdj              0.00     0.00    6.00   60.00    96.00   480.00     8.73     0.17    2.56   1.80  11.90
xvdj              0.00     0.00    4.00    3.00    64.00    24.00    12.57     0.10   15.00  14.71  10.30

and

Code:
Device:         rrqm/s   wrqm/s     r/s     w/s   rsec/s   wsec/s avgrq-sz avgqu-sz   await  svctm  %util
xvdk              0.00     0.09   38.83   28.48  2696.19   228.51    43.45     0.35    5.24   1.29   8.67
xvdk              0.00     0.00  187.00    3.00  5440.00    24.00    28.76     2.65   14.23   5.26 100.00
xvdk              0.00     0.00  330.00   14.00 12592.00   112.00    36.93     3.54   10.15   2.91 100.00
xvdk              0.00     0.00  270.00    0.00  9968.00     0.00    36.92     4.01   14.99   3.70 100.00
xvdk              0.00     0.00  269.00    0.00  8336.00     0.00    30.99     3.27   12.20   3.72 100.00
xvdk              0.00     0.00  249.00    0.00  8496.00     0.00    34.12     3.50   13.47   4.02 100.00
xvdk              0.00     0.00  261.00    0.00  8432.00     0.00    32.31     3.35   13.19   3.83 100.00
xvdk              0.00     0.00  255.00    3.00  8944.00    24.00    34.76     3.56   13.94   3.88 100.00
xvdk              0.00     0.00  255.00   12.00  7872.00    96.00    29.84     3.04   11.38   3.75 100.10
xvdk              0.00     0.00  281.00    0.00  9424.00     0.00    33.54     3.03   10.79   3.56  99.90

These samples are one sec apart. My first guess is that the data sets which are being processed by oracle need reindexing.

Any suggestions would be highly appreciated
 

10 More Discussions You Might Find Interesting

1. HP-UX

Why are these inconsistent?

Hi I am using HP - UX... When I execute the below commands I get these results... Why do i get a different shell value when i run echo $SHELL? # echo $0 ksh # echo $SHELL sh # ps -p$$ PID TTY TIME COMMAND 4013 pts/7 0:00 ksh (2 Replies)
Discussion started by: gurubarancse
2 Replies

2. Shell Programming and Scripting

Inconsistent variable assignment issue

I am using a piece of code which is working fine in some cases and is not working in some other instances. All I am doing is grep'ing and concatenating. Below is a working scenario. $ echo $var1 REPLYFILENAME=Options_NSE1.txt.enc $ FILE_NM=`echo ${var1##*=}`.gz (Take everything after... (1 Reply)
Discussion started by: jamjam10k
1 Replies

3. Shell Programming and Scripting

AWK with inconsistent field layout

Input File: 1234567890123456789012345678901234 ---------------------------------- 11111 XXX xxxxx 6666666666 22222 YYY 2222222222 33333 AAA TTT xxxxx 9999999999 44444 55555 RRR UUU 5555555555 66666 TTT nnnnn 2222222222 77777 XXX ... (4 Replies)
Discussion started by: lordsmiter
4 Replies

4. UNIX for Advanced & Expert Users

inconsistent cat output

Hi I'm executing a menu script in which I `cat a file` but it's giving different output some times. Following is the code fragment taken from my script. while true do cat procs.configured echo ---------separator-------------- sleep 3 done when I execute this code fragment, `cat` outputs... (2 Replies)
Discussion started by: axes
2 Replies

5. Shell Programming and Scripting

ps -eaf gives inconsistent result

The command ps -eaf gives inconsistent result when executed on XYZ server (the command shows 1, 2 or 3 instances of the process. Ideally it should show only 1). I have attached the script file that was executed and the corresponding output. When the same script was executed on ABC server it gave... (1 Reply)
Discussion started by: rebels
1 Replies

6. Shell Programming and Scripting

inconsistent output using date()

I have an input list of: I am just trying to turn this list into its seconds since epoch "+%s" time equivalent. I'm trying to use the following, but every time I run the command it gives me different answers, they are close, but never the same... any clues as to why? xargs -I{} date -j -f... (3 Replies)
Discussion started by: mdlloyd7
3 Replies

7. Shell Programming and Scripting

Inconsistent column printing

Hi, I have a file that has inconsistently numbered columns. Like row1 has 23 columns, and row 2 has 34 columns etc. I would like to re-order the first 8 columns as required and from the 9th column till the end, I would like to print it as it is. I tried to read the re-ordered 8 columns... (7 Replies)
Discussion started by: jacobs.smith
7 Replies

8. Shell Programming and Scripting

Inconsistent behavior when sorting by column

I'm trying to sort the list below, so that multiple instances of the form pass=i_d=j are contiguous. Curiously, the standard sort command to do this works some of the time (it produced the output below) but not all of the time. the command I used was cat fileName | sort -k1.26,1.26n... (3 Replies)
Discussion started by: LeoKSimon
3 Replies

9. UNIX for Advanced & Expert Users

Inconsistent disk information.

On one of the Unix server B.11.31. a disk that has been provisioned from the VMAX, but according to the diskinfo its show coming from clarion. sudo /opt/emc/SYMCLI/bin/symvg sho /dev/vg_<name> Volume Group Name : /dev/vg_<name> Volume Group Type : HP-UX LVM Volume Group State ... (0 Replies)
Discussion started by: kpatel786
0 Replies

10. UNIX for Beginners Questions & Answers

Inconsistent results using sort function

Could you please advise on the following: I have two space-delimited files with 9 and 10 columns, respectively, with exactly the same values in column 1. However, the order of column 1 differs between the two files, so I want to sort both files by column 1, so that I can align them and... (6 Replies)
Discussion started by: aberg
6 Replies
FFCLOCK(2)						      BSD System Calls Manual							FFCLOCK(2)

NAME
ffclock_getcounter, ffclock_getestimate, ffclock_setestimate -- Retrieve feed-forward counter, get and set feed-forward clock estimates LIBRARY
Standard C Library (libc, -lc) SYNOPSIS
#include <sys/timeffc.h> int ffclock_getcounter(ffcounter *ffcount); int ffclock_getestimate(struct ffclock_estimate *cest); int ffclock_setestimate(struct ffclock_estimate *cest); DESCRIPTION
The ffclock is an alternative method to synchronise the system clock. The ffclock implements a feed-forward paradigm and decouples the timestamping and timekeeping kernel functions. This ensures that past clock errors do not affect current timekeeping, an approach radically different from the feedback alternative implemented by the ntpd daemon when adjusting the system clock. The feed-forward approach has demon- strated better performance and higher robustness than a feedback approach when synchronising over the network. In the feed-forward context, a timestamp is a cumulative value of the ticks of the timecounter, which can be converted into seconds by using the feed-forward clock estimates. The ffclock_getcounter() system call allows the calling process to retrieve the current value of the feed-forward counter maintained by the kernel. The ffclock_getestimate() and ffclock_setestimate() system calls allow the caller to get and set the kernel's feed-forward clock parameter estimates respectively. The ffclock_setestimate() system call should be invoked by a single instance of a feed-forward synchronisation dae- mon. The ffclock_getestimate() system call can be called by any process to retrieve the feed-forward clock estimates. The feed-forward approach does not require that the clock estimates be retrieved every time a timestamp is to be converted into seconds. The number of system calls can therefore be greatly reduced if the calling process retrieves the clock estimates from the clock synchronisation daemon instead. The ffclock_getestimate() must be used when the feed-forward synchronisation daemon is not running (see USAGE below). The clock parameter estimates structure pointed to by cest is defined in <sys/timeffc.h> as: struct ffclock_estimate { struct bintime update_time; /* Time of last estimates update. */ ffcounter update_ffcount; /* Counter value at last update. */ ffcounter leapsec_next; /* Counter value of next leap second. */ uint64_t period; /* Estimate of counter period. */ uint32_t errb_abs; /* Bound on absolute clock error [ns]. */ uint32_t errb_rate; /* Bound on counter rate error [ps/s]. */ uint32_t status; /* Clock status. */ int16_t leapsec_total; /* All leap seconds seen so far. */ int8_t leapsec; /* Next leap second (in {-1,0,1}). */ }; Only the super-user may set the feed-forward clock estimates. RETURN VALUES
Upon successful completion, the value 0 is returned; otherwise the value -1 is returned and the global variable errno is set to indicate the error. ERRORS
The following error codes may be set in errno: [EFAULT] The ffcount or cest pointer referenced invalid memory. [EPERM] A user other than the super-user attempted to set the feed-forward clock parameter estimates. USAGE
The feed-forward paradigm enables the definition of specialised clock functions. In its simplest form, ffclock_getcounter() can be used to establish strict order between events or to measure small time intervals very accu- rately with a minimum performance cost. Different methods exist to access absolute time (or "wall-clock time") tracked by the ffclock. The simplest method uses the ffclock sysctl interface kern.ffclock to make the system clock return the ffclock time. The clock_gettime(2) system call can then be used to retrieve the current time seen by the feed-forward clock. Note that this setting affects the entire system and that a feed-forward synchronisation daemon should be running. A less automated method consists of retrieving the feed-forward counter timestamp from the kernel and using the feed-forward clock parameter estimates to convert the timestamp into seconds. The feed-forward clock parameter estimates can be retrieved from the kernel or from the synchronisation daemon directly (preferred). This method allows converting timestamps using different clock models as needed by the applica- tion, while collecting meaningful upper bounds on current clock error. SEE ALSO
date(1), adjtime(2), clock_gettime(2), ctime(3) HISTORY
Feed-forward clock support first appeared in FreeBSD 10.0. AUTHORS
The feed-forward clock support was written by Julien Ridoux <jridoux@unimelb.edu.au> in collaboration with Darryl Veitch <dveitch@unimelb.edu.au> at the University of Melbourne under sponsorship from the FreeBSD Foundation. BSD
November 21, 2011 BSD
All times are GMT -4. The time now is 03:33 AM.
Unix & Linux Forums Content Copyright 1993-2022. All Rights Reserved.
Privacy Policy