Understanding sar "address translation page faults"



We had a problem with an Oracle database instance on S10U3 on a 32G
V890. The database became unresponsive at about 22:00, and we rebooted
the server at 23:00 (database wouldn't shutdown).

We see the following sar output, and the vflt/s numbers spike just as
the database problems begin:

dcpnyc05:~>sar -f /var/adm/sa/sa17 -s 21:00 -e 23:50 -p

SunOS dcpnyc05 5.10 Generic_118833-33 sun4u 07/17/2008

21:00:00 atch/s pgin/s ppgin/s pflt/s vflt/s slock/s
21:05:00 831.69 300.31 857.51 365.03 858.47 0.12
21:10:00 385.74 142.97 408.68 223.62 488.74 0.12
21:15:00 329.95 114.25 195.27 224.31 475.97 0.12
21:20:01 155.23 66.07 111.45 222.96 454.53 0.12
21:25:00 448.27 135.31 465.18 222.77 908.15 0.12
21:30:01 140.41 60.03 98.06 222.29 928.52 0.12
21:35:00 166.11 71.61 131.33 227.57 996.81 0.12
21:40:00 707.88 240.37 750.96 220.37 527.58 0.12
21:45:00 176.71 82.02 145.60 221.20 450.05 0.12
21:50:00 223.11 92.73 184.23 216.75 442.50 0.12
21:55:00 251.98 100.86 218.14 227.75 485.42 0.12
22:00:00 149.45 80.59 131.79 220.05 450.45 0.12
22:05:00 720.57 278.97 783.50 363.48 831.51 0.12
22:10:00 322.64 101.28 191.62 218.93 469.53 0.12
22:15:00 343.92 113.68 297.07 234.01 350984.78 0.12
22:20:00 178.47 43.10 67.37 346.19 460408.41 0.12
22:25:00 173.44 42.63 66.04 329.75 460226.50 0.12
22:30:00 202.78 45.62 76.84 365.96 460261.91 0.12
22:35:00 182.13 54.27 120.96 256.22 459967.38 0.12
22:40:01 103.20 24.85 33.52 241.31 460343.91 0.12
22:45:00 245.99 57.50 191.44 242.99 460024.12 0.12
22:50:00 101.62 26.78 37.43 230.23 459990.97 0.12
22:55:00 125.48 29.42 52.51 243.83 460413.50 0.12
23:00:00 119.30 24.05 32.02 258.81 460429.34 0.13
23:05:00 929.97 18.29 34.87 2049.98 462512.97 0.18
23:30:01 227.42 14.79 43.98 519.04 1227.06 0.00
23:35:00 34.33 0.43 0.54 107.18 228.21 0.00
23:40:00 29.91 0.00 0.00 97.49 206.25 0.00
23:45:00 30.19 0.07 0.15 94.48 207.86 0.00
23:50:00 29.27 0.00 0.00 92.64 200.18 0.00

Average 268.83 78.74 190.84 303.58 165553.48 0.10

The server doesn't appear to be paging to disk during the interval:

dcpnyc05:~>sar -f /var/adm/sa/sa17 -s 21:00 -e 23:50 -w

SunOS dcpnyc05 5.10 Generic_118833-33 sun4u 07/17/2008

21:00:00 swpin/s bswin/s swpot/s bswot/s pswch/s
21:05:00 0.00 0.0 0.00 0.0 14646
21:10:00 0.00 0.0 0.00 0.0 12279
21:15:00 0.00 0.0 0.00 0.0 10198
21:20:01 0.00 0.0 0.00 0.0 8383
21:25:00 0.00 0.0 0.00 0.0 8751
21:30:01 0.00 0.0 0.00 0.0 8768
21:35:00 0.00 0.0 0.00 0.0 15614
21:40:00 0.00 0.0 0.00 0.0 9045
21:45:00 0.00 0.0 0.00 0.0 9136
21:50:00 0.00 0.0 0.00 0.0 9047
21:55:00 0.00 0.0 0.00 0.0 9946
22:00:00 0.00 0.0 0.00 0.0 8524
22:05:00 0.00 0.0 0.00 0.0 13713
22:10:00 0.00 0.0 0.00 0.0 9797
22:15:00 0.00 0.0 0.00 0.0 8831
22:20:00 0.00 0.0 0.00 0.0 9880
22:25:00 0.00 0.0 0.00 0.0 7412
22:30:00 0.00 0.0 0.00 0.0 8543
22:35:00 0.00 0.0 0.00 0.0 9106
22:40:01 0.00 0.0 0.00 0.0 6440
22:45:00 0.00 0.0 0.00 0.0 8661
22:50:00 0.00 0.0 0.00 0.0 6403
22:55:00 0.00 0.0 0.00 0.0 6524
23:00:00 0.00 0.0 0.00 0.0 6494
23:05:00 0.00 0.0 0.00 0.0 6312
23:30:01 0.00 0.0 0.00 0.0 2383
23:35:00 0.00 0.0 0.00 0.0 1735
23:40:00 0.00 0.0 0.00 0.0 1714
23:45:00 0.00 0.0 0.00 0.0 1742
23:50:00 0.00 0.0 0.00 0.0 1712

Average 0.00 0.0 0.00 0.0 8057

According to the manpage, vflt/s is defined as:

vflt/s address translation page
faults per second (valid
page not in memory).

What does this mean, and does it perhaps correlate to our Oracle
database issue?

Any assistance is appreciated.

Cheers,

Paul
.



Relevant Pages

  • Create SharePoint Portal failed.
    ... One mentioned ensuring that SQL Server uses a case ... 13:55:40 Service database server is 'USDC-JOHRIV'. ... Update dbo.propertylist set DisplayName = N'Last name' ...
    (microsoft.public.sharepoint.portalserver)
  • Re: ADO Connection Timeout
    ... to the central server, but you are willing to live with periods where it ... i.e. a local database or even a text file. ... to function until the connection can be restored to the server. ...
    (microsoft.public.data.ado)
  • Web Developers - Happy Hearts And HDTV! - Lockergnome
    ... Certificate on your MSIIS Web server. ... getting data from a database is only half the problem. ... Zend recently started a series about building rock solid code in PHP. ... which provides bulk database conversion. ...
    (freebsd-questions)
  • Re: TNS could not resolve the connect identifier
    ... This database resides on Machine A. ... The Web server is running on Machine B. ... Using tnsping is not as good as using a real connection such as via ... client (note that this is terminology that appears in the 10g R2 ...
    (comp.databases.oracle.server)
  • Config for OLTP system
    ... extrenal disks fo the 60GByte database server. ... IBM Informix Dynamic Server Configuration Parameters ... # BUFFSIZE - OnLine no longer supports this configuration parameter. ...
    (comp.databases.informix)