Negative pages read in chunk

I have some system monitoring that captures pages read from each chunk. This is simply "select fname, pagesread from syschktab" For the most active chunk in the instance this statement returns a large negative number (-1,181,472,262). However an onstat -D returns a larger positive number (3,113,496,195). Other chunks in the instance report the same numbers. This is on version 11.50 Can anyone tell me what is going on here? Is there some magic to turn the syschktab data to the same data returned by onstat -D?

Sounds like a rollover, after a certain amount of incrementing, your positive number becomes negative. j. On 9/12/16 4:11 PM, RAY BURNS wrote: > I have some system monitoring that captures pages read from each chunk. This > is simply "select fname, pagesread from syschktab" > For the most active chunk in the instance this statement returns a large > negative number (-1,181,472,262). However an onstat -D returns a larger > positive number (3,113,496,195). > > Other chunks in the instance report the same numbers. > > This is on version 11.50 > > Can anyone tell me what is going on here? Is there some magic to turn the > syschktab data to the same data returned by onstat -D? > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > >

Unless you're facing a bug which I never noticed, you're probably seeing the nasty effect of some internal counters being too small for some instance activity and uptime (I'm assuming your uptime is good...) The fact that in one place you see a number and on the other you see another may be "only" a display issue. The larget number we can store in 32 bit unsigned is 4.294.967.295. But this same number if interpreted as a signed number is -1. In fact, for a signed number the maximum we can store in 32 bit is:2147483647. If we add 1 we get -2147483648. If we keep adding it will become less negative. The difference between 2147483647 and 1,181,472,262 is roughly 966011386 which is roughly how much you need to add to 2147483648 to get to 3113496195 Confusing? It just means the engine keeps a 64 bit counter but in some places it's showing as an integer (32 bit) which makes it negative. I do recall some fields in sysmaster were changed from INT to INT8 because of this. All this discussion was already explained... my purpose is to try to show you that from my perspective looking at the accumulated values (when then get so big) becomes meaningless... I always recommend to customers to reset the counters on a daily base and if possible store them and keep some history. That allows you to get some pretty graphics, show up anomalous days, or make some pattern change more evident. And this normally also solves the wrap around of the counters. Regards On Mon, Sep 12, 2016 at 10:11 PM, RAY BURNS <ray.burns@velocityglobal.co.nz> wrote: > I have some system monitoring that captures pages read from each chunk. > This > is simply "select fname, pagesread from syschktab" > For the most active chunk in the instance this statement returns a large > negative number (-1,181,472,262). However an onstat -D returns a larger > positive number (3,113,496,195). > > Other chunks in the instance report the same numbers. > > This is on version 11.50 > > Can anyone tell me what is going on here? Is there some magic to turn the > syschktab data to the same data returned by onstat -D? > > > ************************************************************ > ******************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > -- Fernando Nunes Portugal http://informix-technology.blogspot.com My email works... but I don't check it frequently... --001a113dd5267b4dbf053c55df0a

blockquote, div.yahoo_quoted { margin-left: 0 !important; border-left:1px #715FFA solid !important; padding-left:1ex !important; background-color:white !important; } It is the issue of using a signed int vrs an unsigned int. If you place the bit pattern for 3,113,496,195 into a signed 32 bit integer you would get -1,181,471,100, which is roughly what you got. The 'bug' is that pages read is defined to be an integer as there is no unsigned integer SQL data type. Sent from Yahoo Mail for iPad On Monday, September 12, 2016, 3:11 PM, RAY BURNS <ray.burns@velocityglobal.co.nz> wrote: I have some system monitoring that captures pages read from each chunk. This is simply "select fname, pagesread from syschktab" For the most active chunk in the instance this statement returns a large negative number (-1,181,472,262). However an onstat -D returns a larger positive number (3,113,496,195). Other chunks in the instance report the same numbers. This is on version 11.50 Can anyone tell me what is going on here? Is there some magic to turn the syschktab data to the same data returned by onstat -D? ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.

If you want, you could use the following to convert the bit pattern of an integer into an bigint as though the integer were an unsigned integer.... create function itol(val int) returning bigintdefine t bigint; if (val is null) then LET t = 2147483647; LET t = t + 1; elif val > -1 then LET t = val; else LET t = BITAND(val, 2147483647) + 2147483647; LET t = t + 1; end if return t; end function; Madison Pruet Retired and Loving it On Monday, September 12, 2016 5:03 PM, Madison Pruet <madison_pruet@yahoo.com> wrote: #yiv8964042333 blockquote, #yiv8964042333 div.yiv8964042333yahoo_quoted {margin-left:0 !important;border-left:1px #715FFA solid !important;padding-left:1ex !important;background-color:white;} It is the issue of using a signed int vrs an unsigned int. If you place the bit pattern for 3,113,496,195 into a signed 32 bit integer you would get -1,181,471,100, which is roughly what you got. The 'bug' is that pages read is defined to be an integer as there is no unsigned integer SQL data type. Sent from Yahoo Mail for iPad On Monday, September 12, 2016, 3:11 PM, RAY BURNS <ray.burns@velocityglobal.co.nz> wrote: I have some system monitoring that captures pages read from each chunk. This is simply "select fname, pagesread from syschktab" For the most active chunk in the instance this statement returns a large negative number (-1,181,472,262). However an onstat -D returns a larger positive number (3,113,496,195). Other chunks in the instance report the same numbers. This is on version 11.50 Can anyone tell me what is going on here? Is there some magic to turn the syschktab data to the same data returned by onstat -D? ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.

Many thanks for your responses. That has done the trick.