define raw device on Redhat Linux

Folks, IDS11.50 FC7, Redhat Linux5 Our sysadmin is working on Logical volume definition for our new EMC Clarion ... Disk. We saw http://www.informix-zone.com/node/38 is a good paper for getting a Character device ( not Block device) in Suse Linux. Just, want to check if there is more such type of info ... on Redhat ... Thanks, Frank --001485f274e2cfbabd04926ed644

Direct-attached storage on RH 4 and later used /dev/udev/ definition files to recreate at boot time. Don't remember the SAN/PowerPath info off the top of my head. Bob ----- Original Message ----- From: "FRANK" <yunyaoqu@gmail.com> To: ids@iiug.org Sent: Tuesday, October 12, 2010 1:30:59 PM Subject: define raw device on Redhat Linux [21652] Folks, IDS11.50 FC7, Redhat Linux5 Our sysadmin is working on Logical volume definition for our new EMC Clarion ... Disk. We saw http://www.informix-zone.com/node/38 is a good paper for getting a Character device ( not Block device) in Suse Linux. Just, want to check if there is more such type of info ... on Redhat ... Thanks, Frank --001485f274e2cfbabd04926ed644 ******************************************************************************* Forum Note: Use "Reply" to post a response in the discussion forum.

On Tuesday 12 October 2010 19:30:59 FRANK wrote: > Folks, > > IDS11.50 FC7, Redhat Linux5 > > Our sysadmin is working on Logical volume definition for our new EMC > Clarion ... Disk. > We saw http://www.informix-zone.com/node/38 is a good paper for getting > a Character device ( not Block device) in Suse Linux. > Just, want to check if there is more such type of info ... on Redhat ... If O_DIRECT is an option for you: http://dbaspot.com/forums/informix/147617-how-setup-block-devices-o_direct- rhel5.html Thanks, Andreas > > Thanks, > Frank > > --001485f274e2cfbabd04926ed644 > > > *************************************************************************** > **** Forum Note: Use "Reply" to post a response in the discussion forum. > -- Andreas Breitfeld; Informix Development Munich IBM Deutschland Research & Development GmbH; Vorsitzender des Aufsichtsrats: Martin Jetter; Geschaeftsfuehrung: Dirk Wittkopp; Sitz der Gesellschaft: Boeblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294

Thanks, Andreas. I give more detailed on this problem. We are setting up the IDS11.50 FC7 of RedHat Linux 5, using EMC Clariion disk. The Sysadmin defined the raw disk (character device), and Logical Volumes. We did the test to make sure the data would not be buffered(should be written to disk immediately) when the transaction is committed. But, the result is that, after we disabled the disk drivers, IDS keeps running and committing transactions.... So , the committed data is buffered somewhere.... We want to know what is the problem, wrong character device defined? Red Hat and EMC?, or other?...... Basically, How to setup Redhat/EMC raw character device to make sure it does not buffer data when there is a disk problem , so IDS can abort the job immediately. Thanks Frank On Wed, Oct 13, 2010 at 5:08 AM, Andreas Breitfeld <abreitfe@de.ibm.com>wrote: > On Tuesday 12 October 2010 19:30:59 FRANK wrote: > > Folks, > > > > IDS11.50 FC7, Redhat Linux5 > > > > Our sysadmin is working on Logical volume definition for our new EMC > > Clarion ... Disk. > > We saw http://www.informix-zone.com/node/38 is a good paper for getting > > a Character device ( not Block device) in Suse Linux. > > Just, want to check if there is more such type of info ... on Redhat ... > If O_DIRECT is an option for you: > > http://dbaspot.com/forums/informix/147617-how-setup-block-devices-o_direct- > rhel5.html > > Thanks, > Andreas > > > > > Thanks, > > Frank > > > > --001485f274e2cfbabd04926ed644 > > > > > > > *************************************************************************** > > **** Forum Note: Use "Reply" to post a response in the discussion forum. > > > > -- > Andreas Breitfeld; Informix Development Munich > IBM Deutschland Research & Development GmbH; Vorsitzender des > Aufsichtsrats: > Martin Jetter; Geschaeftsfuehrung: Dirk Wittkopp; Sitz der Gesellschaft: > Boeblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294 > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --001485f1e44090f7b8049287541a

Thanks, Bob! I replied with more detailed in another email. Frank On Tue, Oct 12, 2010 at 1:52 PM, rroussey@comcast.net <rroussey@comcast.net>wrote: > Direct-attached storage on RH 4 and later used /dev/udev/ definition files > to > recreate at boot time. Don't remember the SAN/PowerPath info off the top of > my > head. > > Bob > > ----- Original Message ----- > From: "FRANK" <yunyaoqu@gmail.com> > To: ids@iiug.org > Sent: Tuesday, October 12, 2010 1:30:59 PM > Subject: define raw device on Redhat Linux [21652] > > Folks, > > IDS11.50 FC7, Redhat Linux5 > > Our sysadmin is working on Logical volume definition for our new EMC > Clarion ... Disk. > We saw http://www.informix-zone.com/node/38 is a good paper for getting > a Character device ( not Block device) in Suse Linux. > Just, want to check if there is more such type of info ... on Redhat ... > > Thanks, > Frank > > --001485f274e2cfbabd04926ed644 > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --001485f7c3fc6781aa0492875979

Frank, The problem is not your RAW device definitions, it is your understanding of what Informix is doing and how it does transactions. Remember that Informix ONLY works on data pages that are in its cache, so as long as the data that is being modified by your transactions is already in cache (or would be on a new page and so only written initially to cache) Informix will not be modifying the chunks on disk and will not realize that the disk has gone bad. Dirty pages are flushed to disk when the lru_min/max_dirty thresholds are reached or during or after a checkpoint. There is one exception. Logical and physical log pages are immediately written to disk. You could try destroying the raw partition or physically removing the disk drive for the logical logs and see what happens, but that's not going to get you quicker responses for the other chunks. Also, is it possible that the engine is actually still able to write to the device after you disabled the device drivers? Perhaps the OS is maintaining the path to the raw device even with the device driver removed because a process has the device open. Your test may not be valid. Try physically removing or powering down the drive itself. Art Art S. Kagel Advanced DataTools (www.advancedatatools.com) IIUG Board of Directors (art@iiug.org) Disclaimer: Please keep in mind that my own opinions are my own opinions and do not reflect on my employer, Advanced DataTools, the IIUG, nor any other organization with which I am associated either explicitly, implicitly, or by inference. Neither do those opinions reflect those of other individuals affiliated with any entity with which I am affiliated nor those of the entities themselves. On Wed, Oct 13, 2010 at 6:44 PM, FRANK <yunyaoqu@gmail.com> wrote: > Thanks, Andreas. > I give more detailed on this problem. > > We are setting up the IDS11.50 FC7 of RedHat Linux 5, using EMC Clariion > disk. > The Sysadmin defined the raw disk (character device), and Logical Volumes. > > We did the test to make sure the data would not be buffered(should be > written to disk immediately) when the transaction is committed. > > But, the result is that, after we disabled the disk drivers, IDS keeps > running and committing transactions.... So , the committed data is > buffered somewhere.... > > We want to know what is the problem, wrong character device defined? Red > Hat and EMC?, or other?...... > > Basically, How to setup Redhat/EMC raw character device to make sure it > does not buffer data when there is a disk problem , so IDS can abort the > job immediately. > > Thanks > Frank > > On Wed, Oct 13, 2010 at 5:08 AM, Andreas Breitfeld <abreitfe@de.ibm.com > >wrote: > > > On Tuesday 12 October 2010 19:30:59 FRANK wrote: > > > Folks, > > > > > > IDS11.50 FC7, Redhat Linux5 > > > > > > Our sysadmin is working on Logical volume definition for our new EMC > > > Clarion ... Disk. > > > We saw http://www.informix-zone.com/node/38 is a good paper for > getting > > > a Character device ( not Block device) in Suse Linux. > > > Just, want to check if there is more such type of info ... on Redhat > ... > > If O_DIRECT is an option for you: > > > > > http://dbaspot.com/forums/informix/147617-how-setup-block-devices-o_direct- > > rhel5.html > > > > Thanks, > > Andreas > > > > > > > > Thanks, > > > Frank > > > > > > --001485f274e2cfbabd04926ed644 > > > > > > > > > > > > *************************************************************************** > > > **** Forum Note: Use "Reply" to post a response in the discussion > forum. > > > > > > > -- > > Andreas Breitfeld; Informix Development Munich > > IBM Deutschland Research & Development GmbH; Vorsitzender des > > Aufsichtsrats: > > Martin Jetter; Geschaeftsfuehrung: Dirk Wittkopp; Sitz der Gesellschaft: > > Boeblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294 > > > > > > > > > > ******************************************************************************* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > --001485f1e44090f7b8049287541a > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --0016361e81deaa5bc10492883076

Thanks a lot, Art! Sorry not gave the description clearly. This is a brand new installation and testing. (1) created a new db root chunk ( the only one we created) pointed to the raw character device defined by the system administrator. It was the only chunk we defined for the instance that was used for every thing : root dbs, logical/physical log, data ....... (2) created a new table in the chunk, started a looping script that just kept inserting new row into the table (3) pulled out the two disk driver cables( physically cut the disk connection, so NO DISK available now!) that bridged Disk to EMC Powerpath. The RedHat connected to EMC powerpath. IDS was not impacted, stayed online as normal. (4) we saw the loop script job paused for around 20 seconds...., then resumed doing the inserting job perfectly...... (5)IDS keep working continuously without any sense of disk failure during the whole procedure. ( Logical log completed one by one.....) So, we think the IDS was cheated somehow... and data was taken by a Middleman..... So, we are looking for the Middleman... Thanks, Frank On Wed, Oct 13, 2010 at 7:45 PM, Art Kagel <art.kagel@gmail.com> wrote: > Frank, > > The problem is not your RAW device definitions, it is your understanding of > what Informix is doing and how it does transactions. Remember that Informix > ONLY works on data pages that are in its cache, so as long as the data that > is being modified by your transactions is already in cache (or would be on > a > new page and so only written initially to cache) Informix will not be > modifying the chunks on disk and will not realize that the disk has gone > bad. Dirty pages are flushed to disk when the lru_min/max_dirty thresholds > are reached or during or after a checkpoint. There is one exception. > Logical and physical log pages are immediately written to disk. You could > try destroying the raw partition or physically removing the disk drive for > the logical logs and see what happens, but that's not going to get you > quicker responses for the other chunks. > > Also, is it possible that the engine is actually still able to write to the > device after you disabled the device drivers? Perhaps the OS is maintaining > the path to the raw device even with the device driver removed because a > process has the device open. Your test may not be valid. Try physically > removing or powering down the drive itself. > > Art > > Art S. Kagel > Advanced DataTools (www.advancedatatools.com) > IIUG Board of Directors (art@iiug.org) > > Disclaimer: Please keep in mind that my own opinions are my own opinions > and > do not reflect on my employer, Advanced DataTools, the IIUG, nor any other > organization with which I am associated either explicitly, implicitly, or > by > inference. Neither do those opinions reflect those of other individuals > affiliated with any entity with which I am affiliated nor those of the > entities themselves. > > On Wed, Oct 13, 2010 at 6:44 PM, FRANK <yunyaoqu@gmail.com> wrote: > > > Thanks, Andreas. > > I give more detailed on this problem. > > > > We are setting up the IDS11.50 FC7 of RedHat Linux 5, using EMC Clariion > > disk. > > The Sysadmin defined the raw disk (character device), and Logical > Volumes. > > > > We did the test to make sure the data would not be buffered(should be > > written to disk immediately) when the transaction is committed. > > > > But, the result is that, after we disabled the disk drivers, IDS keeps > > running and committing transactions.... So , the committed data is > > buffered somewhere.... > > > > We want to know what is the problem, wrong character device defined? Red > > Hat and EMC?, or other?...... > > > > Basically, How to setup Redhat/EMC raw character device to make sure it > > does not buffer data when there is a disk problem , so IDS can abort the > > job immediately. > > > > Thanks > > Frank > > > > On Wed, Oct 13, 2010 at 5:08 AM, Andreas Breitfeld <abreitfe@de.ibm.com > > >wrote: > > > > > On Tuesday 12 October 2010 19:30:59 FRANK wrote: > > > > Folks, > > > > > > > > IDS11.50 FC7, Redhat Linux5 > > > > > > > > Our sysadmin is working on Logical volume definition for our new EMC > > > > Clarion ... Disk. > > > > We saw http://www.informix-zone.com/node/38 is a good paper for > > getting > > > > a Character device ( not Block device) in Suse Linux. > > > > Just, want to check if there is more such type of info ... on Redhat > > ... > > > If O_DIRECT is an option for you: > > > > > > > > > http://dbaspot.com/forums/informix/147617-how-setup-block-devices-o_direct- > > > rhel5.html > > > > > > Thanks, > > > Andreas > > > > > > > > > > > Thanks, > > > > Frank > > > > > > > > --001485f274e2cfbabd04926ed644 > > > > > > > > > > > > > > > > > > *************************************************************************** > > > > **** Forum Note: Use "Reply" to post a response in the discussion > > forum. > > > > > > > > > > -- > > > Andreas Breitfeld; Informix Development Munich > > > IBM Deutschland Research & Development GmbH; Vorsitzender des > > > Aufsichtsrats: > > > Martin Jetter; Geschaeftsfuehrung: Dirk Wittkopp; Sitz der > Gesellschaft: > > > Boeblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294 > > > > > > > > > > > > > > > > > > ******************************************************************************* > > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > > > > > --001485f1e44090f7b8049287541a > > > > > > > > > > ******************************************************************************* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > --0016361e81deaa5bc10492883076 > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > > --001485f03cac6f58c504928a6acd

The "Middleman " is still on the loose.... IDS could complete Check points , flush data... to disk successfully without disks available! It is strongly suspected that the "guy" is hidden in the "raw character devices".... created by......, agree? Good news: The Boss has rewarded $10,000 for any tips leading to the capture of the "Middleman"! Good luck! Frank On Wed, Oct 13, 2010 at 10:25 PM, FRANK <yunyaoqu@gmail.com> wrote: > Thanks a lot, Art! > > Sorry not gave the description clearly. > > This is a brand new installation and testing. > > (1) created a new db root chunk ( the only one we created) pointed to the > raw character device defined by the system administrator. It was the only > chunk we defined for the instance that was used for every thing : root > dbs, logical/physical log, data ....... > (2) created a new table in the chunk, started a looping script that > just kept inserting new row into the table > (3) pulled out the two disk driver cables( physically cut the disk > connection, so NO DISK available now!) that bridged Disk to EMC > Powerpath. The RedHat connected to EMC powerpath. IDS was not impacted, > stayed online as normal. > (4) we saw the loop script job paused for around 20 seconds...., then > resumed doing the inserting job perfectly...... > (5)IDS keep working continuously without any sense of disk failure during > the whole procedure. ( Logical log completed one by one.....) > > So, we think the IDS was cheated somehow... and data was taken by a > Middleman..... So, we are looking for the Middleman... > > Thanks, > Frank > > On Wed, Oct 13, 2010 at 7:45 PM, Art Kagel <art.kagel@gmail.com> wrote: > > > Frank, > > > > The problem is not your RAW device definitions, it is your understanding > of > > what Informix is doing and how it does transactions. Remember that > Informix > > ONLY works on data pages that are in its cache, so as long as the data > that > > is being modified by your transactions is already in cache (or would be > on > > a > > new page and so only written initially to cache) Informix will not be > > modifying the chunks on disk and will not realize that the disk has gone > > bad. Dirty pages are flushed to disk when the lru_min/max_dirty > thresholds > > are reached or during or after a checkpoint. There is one exception. > > Logical and physical log pages are immediately written to disk. You could > > try destroying the raw partition or physically removing the disk drive > for > > the logical logs and see what happens, but that's not going to get you > > quicker responses for the other chunks. > > > > Also, is it possible that the engine is actually still able to write to > the > > device after you disabled the device drivers? Perhaps the OS is > maintaining > > the path to the raw device even with the device driver removed because a > > process has the device open. Your test may not be valid. Try physically > > removing or powering down the drive itself. > > > > Art > > > > Art S. Kagel > > Advanced DataTools (www.advancedatatools.com) > > IIUG Board of Directors (art@iiug.org) > > > > Disclaimer: Please keep in mind that my own opinions are my own opinions > > and > > do not reflect on my employer, Advanced DataTools, the IIUG, nor any > other > > organization with which I am associated either explicitly, implicitly, or > > by > > inference. Neither do those opinions reflect those of other individuals > > affiliated with any entity with which I am affiliated nor those of the > > entities themselves. > > > > On Wed, Oct 13, 2010 at 6:44 PM, FRANK <yunyaoqu@gmail.com> wrote: > > > > > Thanks, Andreas. > > > I give more detailed on this problem. > > > > > > We are setting up the IDS11.50 FC7 of RedHat Linux 5, using EMC > Clariion > > > disk. > > > The Sysadmin defined the raw disk (character device), and Logical > > Volumes. > > > > > > We did the test to make sure the data would not be buffered(should be > > > written to disk immediately) when the transaction is committed. > > > > > > But, the result is that, after we disabled the disk drivers, IDS keeps > > > running and committing transactions.... So , the committed data is > > > buffered somewhere.... > > > > > > We want to know what is the problem, wrong character device defined? > Red > > > Hat and EMC?, or other?...... > > > > > > Basically, How to setup Redhat/EMC raw character device to make sure it > > > does not buffer data when there is a disk problem , so IDS can abort > the > > > job immediately. > > > > > > Thanks > > > Frank > > > > > > On Wed, Oct 13, 2010 at 5:08 AM, Andreas Breitfeld < > abreitfe@de.ibm.com > > > >wrote: > > > > > > > On Tuesday 12 October 2010 19:30:59 FRANK wrote: > > > > > Folks, > > > > > > > > > > IDS11.50 FC7, Redhat Linux5 > > > > > > > > > > Our sysadmin is working on Logical volume definition for our new > EMC > > > > > Clarion ... Disk. > > > > > We saw http://www.informix-zone.com/node/38 is a good paper for > > > getting > > > > > a Character device ( not Block device) in Suse Linux. > > > > > Just, want to check if there is more such type of info ... on > Redhat > > > ... > > > > If O_DIRECT is an option for you: > > > > > > > > > > > > > > http://dbaspot.com/forums/informix/147617-how-setup-block-devices-o_direct- > > > > rhel5.html > > > > > > > > Thanks, > > > > Andreas > > > > > > > > > > > > > > Thanks, > > > > > Frank > > > > > > > > > > --001485f274e2cfbabd04926ed644 > > > > > > > > > > > > > > > > > > > > > > > > > *************************************************************************** > > > > > **** Forum Note: Use "Reply" to post a response in the discussion > > > forum. > > > > > > > > > > > > > -- > > > > Andreas Breitfeld; Informix Development Munich > > > > IBM Deutschland Research & Development GmbH; Vorsitzender des > > > > Aufsichtsrats: > > > > Martin Jetter; Geschaeftsfuehrung: Dirk Wittkopp; Sitz der > > Gesellschaft: > > > > Boeblingen; Registergericht: Amtsgericht Stuttgart, HRB 243294 > > > > > > > > > > > > > > > > > > > > > > > > > > > > ******************************************************************************* > > > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > > > > > > > > > --001485f1e44090f7b8049287541a > > > > > > > > > > > > > > > > > > ******************************************************************************* > > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > > > > > --0016361e81deaa5bc10492883076 > > > > > > > > > > ******************************************************************************* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > --001485f03cac6f58c504928a6acd > > > > ******************