Cooked vs Raw (DBA vs Sysadmin)
A DBA asked who should decide whether IDS 10 on Red Hat uses cooked files or raw devices, after a new sysadmin pushed for cooked (ext3) files so all servers would look alike and he could "see" and recover the data. Replies largely backed raw: the DBA should own disk layout, raw avoids filesystem buffering (writes the engine thinks are on disk may be lost in a crash), and cooked files can be accidentally deleted; admins gain little since online file copies aren't valid backups and df is misleading. Others countered that on striped/SAN storage (e.g. EMC with LUN-level copies) benchmarked performance was comparable. It's an opinion thread with no decision recorded.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
Just looking for some opinions regarding who should decide if and IDS 10.00.FC5 (RH 4.0 ES) database should run on cooked files or raw partitions. I am currently in a tug of war with the "new" sysadmin. I use to DBA and sysadmin our database boxes but somewhere along the line that responsibility was reassigned at some point. The only argument the sysadmin has been able to make for cooked files is that all our new server machines DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that he would be unable to see the database files on a raw partition. He believes that the system is less prone to failure, and would it would be easier to recover the database if we used cooked files. Comments? -Peter
Hmmm... none of the users will care how the system "looks" when performance is poor. And the sysadmin does not need to "see" the database. The sysad can own the system. but the DBA should "own" the disks and determine what is laid where to ensure optimum performance. On Aug 22, 2006, at 11:17 AM, Peter J Diaz de Leon wrote: > > Just looking for some opinions regarding who should decide if and IDS > 10.00.FC5 (RH 4.0 ES) database > should run on cooked files or raw partitions. I am currently in a tug > of war with the "new" sysadmin. I use to > DBA and sysadmin our database boxes but somewhere along the line that > responsibility was reassigned at some > point. The only argument the sysadmin has been able to make for cooked > files is that all our new server machines > DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that > he would be unable to see the > database files on a raw partition. He believes that the system is less > prone to failure, and would it would be easier > to recover the database if we used cooked files. > > Comments? > > -Peter > > > ********************************************************************** > ********* > Forum Note: Use "Reply" to post a response in the discussion forum. >
And that´s not it all. I would say that anyone can "delete" a cooked. Raw is less easy to use but more safely to work. >From: "Christine Normile" <christinenormile@mac.com> >Reply-To: ids@iiug.org >To: ids@iiug.org >Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7323] >Date: Tue, 22 Aug 2006 13:05:45 -0400 (EDT) > >Hmmm... none of the users will care how the system "looks" when >performance is poor. And the sysadmin does not need to "see" the >database. The sysad can own the system. but the DBA should "own" the >disks and determine what is laid where to ensure optimum performance. >On Aug 22, 2006, at 11:17 AM, Peter J Diaz de Leon wrote: > > > > > Just looking for some opinions regarding who should decide if and IDS > > 10.00.FC5 (RH 4.0 ES) database > > should run on cooked files or raw partitions. I am currently in a tug > > of war with the "new" sysadmin. I use to > > DBA and sysadmin our database boxes but somewhere along the line that > > responsibility was reassigned at some > > point. The only argument the sysadmin has been able to make for cooked > > files is that all our new server machines > > DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that > > he would be unable to see the > > database files on a raw partition. He believes that the system is less > > prone to failure, and would it would be easier > > to recover the database if we used cooked files. > > > > Comments? > > > > -Peter > > > > > > ********************************************************************** > > ********* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > >******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. >
Also to mention that Cooked goes through the file system. What the databas= e thinks it has written to disk may still be in the buffer when something b= reaks. Not the case with Raw. It's an argument I've had mutiple times and won or lost depending on who 'o= wns' the disks. I had a 1.2TB system that I lost the argument on - we woun= d up benchmarking raw vs stripped/cooked (4 disk stripe sets). Performance= was comparable. j. >From: =3D?ISO8859-1?Q?Andr=3DC3=3DA9_Luiz_Rufino?=3D <andre_rufino@msn.com= > >Date: 2006/08/22 Tue PM 12:10:50 CDT >To: ids@iiug.org >Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7324] > >And that=B4s not it all. I would say that anyone can "delete" a cooked. Ra= w is=20 >less easy to use but more safely to work.=20 > >>From: "Christine Normile" <christinenormile@mac.com>=20 >>Reply-To: ids@iiug.org=20 >>To: ids@iiug.org=20 >>Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7323]=20 >>Date: Tue, 22 Aug 2006 13:05:45 -0400 (EDT)=20 >>=20 >>Hmmm... none of the users will care how the system "looks" when=20 >>performance is poor. And the sysadmin does not need to "see" the=20 >>database. The sysad can own the system. but the DBA should "own" the=20 >>disks and determine what is laid where to ensure optimum performance.=20 >>On Aug 22, 2006, at 11:17 AM, Peter J Diaz de Leon wrote:=20 >>=20 >> >=20 >> > Just looking for some opinions regarding who should decide if and IDS= =20 >> > 10.00.FC5 (RH 4.0 ES) database=20 >> > should run on cooked files or raw partitions. I am currently in a tug= =20 >> > of war with the "new" sysadmin. I use to=20 >> > DBA and sysadmin our database boxes but somewhere along the line that= =20 >> > responsibility was reassigned at some=20 >> > point. The only argument the sysadmin has been able to make for cooked= =20 >> > files is that all our new server machines=20 >> > DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that= =20 >> > he would be unable to see the=20 >> > database files on a raw partition. He believes that the system is less= =20 >> > prone to failure, and would it would be easier=20 >> > to recover the database if we used cooked files.=20 >> >=20 >> > Comments?=20 >> >=20 >> > -Peter=20 >> >=20 >> >=20 >> > **********************************************************************= =20 >> > *********=20 >> > Forum Note: Use "Reply" to post a response in the discussion forum.=20 >> >=20 >>=20 >>=20 > >>*************************************************************************= ******=20 >> Forum Note: Use "Reply" to post a response in the discussion forum.=20 >>=20 > > >**************************************************************************= *****=20 > Forum Note: Use "Reply" to post a response in the discussion forum.=20
This may sound snotty, but a sysadmin that knows his stuff would NOT be complaining about "unable to see" database files. Bob Roussey Unix / Informix Administration Spirit Airlines Robert.Roussey@SpiritAir.com -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of vze2qjg5@verizon.net Sent: Tuesday, August 22, 2006 1:27 PM To: ids@iiug.org Subject: Re: Re: Cooked vs Raw (DBA vs Sysadmin) [7325] Also to mention that Cooked goes through the file system. What the databas= e thinks it has written to disk may still be in the buffer when something b= reaks. Not the case with Raw. It's an argument I've had mutiple times and won or lost depending on who 'o= wns' the disks. I had a 1.2TB system that I lost the argument on - we woun= d up benchmarking raw vs stripped/cooked (4 disk stripe sets). Performance= was comparable. j. >From: =3D?ISO8859-1?Q?Andr=3DC3=3DA9_Luiz_Rufino?=3D <andre_rufino@msn.com= > >Date: 2006/08/22 Tue PM 12:10:50 CDT >To: ids@iiug.org >Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7324] > >And that=B4s not it all. I would say that anyone can "delete" a cooked. Ra= w is=20 >less easy to use but more safely to work.=20 > >>From: "Christine Normile" <christinenormile@mac.com>=20 >>Reply-To: ids@iiug.org=20 >>To: ids@iiug.org=20 >>Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7323]=20 >>Date: Tue, 22 Aug 2006 13:05:45 -0400 (EDT)=20 >>=20 >>Hmmm... none of the users will care how the system "looks" when=20 >>performance is poor. And the sysadmin does not need to "see" the=20 >>database. The sysad can own the system. but the DBA should "own" the=20 >>disks and determine what is laid where to ensure optimum performance.=20 >>On Aug 22, 2006, at 11:17 AM, Peter J Diaz de Leon wrote:=20 >>=20 >> >=20 >> > Just looking for some opinions regarding who should decide if and IDS= =20 >> > 10.00.FC5 (RH 4.0 ES) database=20 >> > should run on cooked files or raw partitions. I am currently in a tug= =20 >> > of war with the "new" sysadmin. I use to=20 >> > DBA and sysadmin our database boxes but somewhere along the line that= =20 >> > responsibility was reassigned at some=20 >> > point. The only argument the sysadmin has been able to make for cooked= =20 >> > files is that all our new server machines=20 >> > DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that= =20 >> > he would be unable to see the=20 >> > database files on a raw partition. He believes that the system is less= =20 >> > prone to failure, and would it would be easier=20 >> > to recover the database if we used cooked files.=20 >> >=20 >> > Comments?=20 >> >=20 >> > -Peter=20 >> >=20 >> >=20 >> > **********************************************************************= =20 >> > *********=20 >> > Forum Note: Use "Reply" to post a response in the discussion forum.=20 >> >=20 >>=20 >>=20 > >>********************************************************************** ***= ******=20 >> Forum Note: Use "Reply" to post a response in the discussion forum.=20 >>=20 > > >*********************************************************************** ***= *****=20 > Forum Note: Use "Reply" to post a response in the discussion forum.=20 ************************************************************************ ******* Forum Note: Use "Reply" to post a response in the discussion forum.
The sysadmin wants the cooked files on an EXT3 partition in case that matters. -Peter vze2qjg5@verizon.net wrote: > Also to mention that Cooked goes through the file system. What the databas= > e thinks it has written to disk may still be in the buffer when something b= > reaks. Not the case with Raw. > > It's an argument I've had mutiple times and won or lost depending on who 'o= > wns' the disks. I had a 1.2TB system that I lost the argument on - we woun= > d up benchmarking raw vs stripped/cooked (4 disk stripe sets). Performance= > was comparable. > > j. > >
Jack, I have lost a few arguments on this as well, especially with what is now called "Fast SAN". The performance of those drives was so close to that of the raw disks that the client decided to go "flat". Their Disaster Recovery Plan was using EMC to copy the LUNs containing the database and its programming to another set of disks using what was called a "SAN-to-SAN Copy". Whether it needed to be that way for that solution to work I don't recall but it was a pretty nice setup. Take care. Clifton -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of vze2qjg5@verizon.net Sent: Tuesday, August 22, 2006 12:27 PM To: ids@iiug.org Subject: Re: Re: Cooked vs Raw (DBA vs Sysadmin) [7325] Also to mention that Cooked goes through the file system. What the databas= e thinks it has written to disk may still be in the buffer when something b= reaks. Not the case with Raw. It's an argument I've had mutiple times and won or lost depending on who 'o= wns' the disks. I had a 1.2TB system that I lost the argument on - we woun= d up benchmarking raw vs stripped/cooked (4 disk stripe sets). Performance= was comparable. j. >From: =3D?ISO8859-1?Q?Andr=3DC3=3DA9_Luiz_Rufino?=3D <andre_rufino@msn.com= > >Date: 2006/08/22 Tue PM 12:10:50 CDT >To: ids@iiug.org >Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7324] > >And that=B4s not it all. I would say that anyone can "delete" a cooked. Ra= w is=20 >less easy to use but more safely to work.=20 > >>From: "Christine Normile" <christinenormile@mac.com>=20 >>Reply-To: ids@iiug.org=20 >>To: ids@iiug.org=20 >>Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7323]=20 >>Date: Tue, 22 Aug 2006 13:05:45 -0400 (EDT)=20 >>=20 >>Hmmm... none of the users will care how the system "looks" when=20 >>performance is poor. And the sysadmin does not need to "see" the=20 >>database. The sysad can own the system. but the DBA should "own" the=20 >>disks and determine what is laid where to ensure optimum performance.=20 >>On Aug 22, 2006, at 11:17 AM, Peter J Diaz de Leon wrote:=20 >>=20 >> >=20 >> > Just looking for some opinions regarding who should decide if and IDS= =20 >> > 10.00.FC5 (RH 4.0 ES) database=20 >> > should run on cooked files or raw partitions. I am currently in a tug= =20 >> > of war with the "new" sysadmin. I use to=20 >> > DBA and sysadmin our database boxes but somewhere along the line that= =20 >> > responsibility was reassigned at some=20 >> > point. The only argument the sysadmin has been able to make for cooked= =20 >> > files is that all our new server machines=20 >> > DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that= =20 >> > he would be unable to see the=20 >> > database files on a raw partition. He believes that the system is less= =20 >> > prone to failure, and would it would be easier=20 >> > to recover the database if we used cooked files.=20 >> >=20 >> > Comments?=20 >> >=20 >> > -Peter=20 >> >=20 >> >=20 >> > **********************************************************************= =20 >> > *********=20 >> > Forum Note: Use "Reply" to post a response in the discussion forum.=20 >> >=20 >>=20 >>=20 > >>*************************************************************************= ******=20 >> Forum Note: Use "Reply" to post a response in the discussion forum.=20 >>=20 > > >**************************************************************************= *****=20 > Forum Note: Use "Reply" to post a response in the discussion forum.=20 **************************************************************************** *** Forum Note: Use "Reply" to post a response in the discussion forum.
Should have clarified - the system in question was in fact an EMC SAN. EMC's BCP utility is very nice for backup and DRP planning - saw it work very nicely in an orrible environment. j. >From: Clifton Bean <cbean@ctm.com> >Date: 2006/08/22 Tue PM 12:41:29 CDT >To: ids@iiug.org >Subject: RE: Re: Cooked vs Raw (DBA vs Sysadmin) [7328] > >Jack, > >I have lost a few arguments on this as well, especially with what is now >called "Fast SAN". The performance of those drives was so close to that of >the raw disks that the client decided to go "flat". > >Their Disaster Recovery Plan was using EMC to copy the LUNs containing the >database and its programming to another set of disks using what was called a >"SAN-to-SAN Copy". Whether it needed to be that way for that solution to >work I don't recall but it was a pretty nice setup. > >Take care. >Clifton > >-----Original Message----- >From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of >vze2qjg5@verizon.net >Sent: Tuesday, August 22, 2006 12:27 PM >To: ids@iiug.org >Subject: Re: Re: Cooked vs Raw (DBA vs Sysadmin) [7325] > >Also to mention that Cooked goes through the file system. What the databas= >e thinks it has written to disk may still be in the buffer when something b= > >reaks. Not the case with Raw. > >It's an argument I've had mutiple times and won or lost depending on who 'o= > >wns' the disks. I had a 1.2TB system that I lost the argument on - we woun= >d up benchmarking raw vs stripped/cooked (4 disk stripe sets). Performance= >was comparable. > >j. > >>From: =3D?ISO8859-1?Q?Andr=3DC3=3DA9_Luiz_Rufino?=3D <andre_rufino@msn.com= > >> >>Date: 2006/08/22 Tue PM 12:10:50 CDT >>To: ids@iiug.org >>Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7324] > >> >>And that=B4s not it all. I would say that anyone can "delete" a cooked. Ra= > >w is=20 >>less easy to use but more safely to work.=20 >> >>>From: "Christine Normile" <christinenormile@mac.com>=20 >>>Reply-To: ids@iiug.org=20 >>>To: ids@iiug.org=20 >>>Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7323]=20 >>>Date: Tue, 22 Aug 2006 13:05:45 -0400 (EDT)=20 >>>=20 >>>Hmmm... none of the users will care how the system "looks" when=20 >>>performance is poor. And the sysadmin does not need to "see" the=20 >>>database. The sysad can own the system. but the DBA should "own" the=20 >>>disks and determine what is laid where to ensure optimum performance.=20 >>>On Aug 22, 2006, at 11:17 AM, Peter J Diaz de Leon wrote:=20 >>>=20 >>> >=20 >>> > Just looking for some opinions regarding who should decide if and IDS= >=20 >>> > 10.00.FC5 (RH 4.0 ES) database=20 >>> > should run on cooked files or raw partitions. I am currently in a tug= >=20 >>> > of war with the "new" sysadmin. I use to=20 >>> > DBA and sysadmin our database boxes but somewhere along the line that= >=20 >>> > responsibility was reassigned at some=20 >>> > point. The only argument the sysadmin has been able to make for cooked= > >=20 >>> > files is that all our new server machines=20 >>> > DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that= > >=20 >>> > he would be unable to see the=20 >>> > database files on a raw partition. He believes that the system is less= > >=20 >>> > prone to failure, and would it would be easier=20 >>> > to recover the database if we used cooked files.=20 >>> >=20 >>> > Comments?=20 >>> >=20 >>> > -Peter=20 >>> >=20 >>> >=20 >>> > **********************************************************************= > >=20 >>> > *********=20 >>> > Forum Note: Use "Reply" to post a response in the discussion forum.=20 >>> >=20 >>>=20 >>>=20 >> >>>*************************************************************************= > >******=20 >>> Forum Note: Use "Reply" to post a response in the discussion forum.=20 >>>=20 >> >> >>**************************************************************************= > >*****=20 >> Forum Note: Use "Reply" to post a response in the discussion forum.=20 > >**************************************************************************** >*** >Forum Note: Use "Reply" to post a response in the discussion forum. > > >******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum.
Hi, my opinion on this is: the admin does not really benefit from having the database as files in the filesystem: - he cannot make a nice backup from just copying the files while the db is online - he cannot see if the files are almost full ( df command is useless, since informix occupies the whole space initially). - he might suffer a performance loss in comparison to raw disks (depending on the hardware, of course). The only thing I do sometimes (but only on development systems) is to have a cooked database chunk to be able to make a fast copy of it. But it always depends on the personal opinion of the admin ... Marcus ----- Original Message ----- From: "Christine Normile" <christinenormile@mac.com> To: <ids@iiug.org> Sent: Tuesday, August 22, 2006 7:05 PM Subject: Re: Cooked vs Raw (DBA vs Sysadmin) [7323] > > Hmmm... none of the users will care how the system "looks" when > performance is poor. And the sysadmin does not need to "see" the > database. The sysad can own the system. but the DBA should "own" the > disks and determine what is laid where to ensure optimum performance. > On Aug 22, 2006, at 11:17 AM, Peter J Diaz de Leon wrote: > > > > > Just looking for some opinions regarding who should decide if and IDS > > 10.00.FC5 (RH 4.0 ES) database > > should run on cooked files or raw partitions. I am currently in a tug > > of war with the "new" sysadmin. I use to > > DBA and sysadmin our database boxes but somewhere along the line that > > responsibility was reassigned at some > > point. The only argument the sysadmin has been able to make for cooked > > files is that all our new server machines > > DB, WEB, JAVA APP, REPORT will look the same. He is uncomfortable that > > he would be unable to see the > > database files on a raw partition. He believes that the system is less > > prone to failure, and would it would be easier > > to recover the database if we used cooked files. > > > > Comments? > > > > -Peter > > > > > > ********************************************************************** > > ********* > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > **************************************************************************** *** > Forum Note: Use "Reply" to post a response in the discussion forum.