NetBSD Problem Report #26741

Received: (qmail 28060 invoked by uid 605); 23 Aug 2004 12:23:29 -0000
Message-Id: <200408231223.i7NCNMkh003483@colwyn.zhadum.de>
Date: Mon, 23 Aug 2004 14:23:22 +0200 (CEST)
From: Matthias Scheler <tron@zhadum.de>
Sender: gnats-bugs-owner@NetBSD.org
Reply-To: tron@zhadum.de
To: gnats-bugs@gnats.NetBSD.org
Subject: "raidctl -S" doesn't always work as expected
X-Send-Pr-Version: 3.95

>Number:         26741
>Category:       bin
>Synopsis:       "raidctl -S" doesn't always work as expected
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    oster
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Aug 23 12:24:00 +0000 2004
>Closed-Date:    
>Last-Modified:  Fri Aug 27 15:58:00 +0000 2004
>Originator:     
>Release:        NetBSD 2.0_BETA
>Organization:
Matthias Scheler                                  http://scheler.de/~matthias/
>Environment:
System: NetBSD colwyn.zhadum.de 2.0_BETA NetBSD 2.0_BETA (COLWYN) #0: Mon Jul 26 12:22:36 CEST 2004 tron@colwyn.zhadum.de:/src/sys/compile/COLWYN i386
Architecture: i386
Machine: i386
>Description:
My system suffered death by administration a few minutes ago and is now
rewriting the parity on the RAIDframe mirror "raid0". "raidctl -S raid0"
works nicely here:

tron@colwyn:~>raidctl -S raid0
Reconstruction is 100% complete.
Parity Re-write is 0% complete.
Copyback is 100% complete.
Parity Re-write status:
  4% |*                                      | ETA: 01:41:31 /

But although the parity of the other mirror is also dirty ...

tron@colwyn:~>raidctl -s raid1
[...]
Parity status: DIRTY
Reconstruction is 100% complete.
Parity Re-write is 100% complete.
Copyback is 100% complete.

... "raidctl -S raid1" returns immediately as if everything is allright:

tron@colwyn:~>time raidctl -S raid1
Reconstruction is 100% complete.
Parity Re-write is 100% complete.
Copyback is 100% complete.
raidctl -S raid1  0.01s user 0.00s system 354% cpu 0.003 total

It should at least print a warning that the mirror isn't in sync.

>How-To-Repeat:
Hard crash a system with two RAIDframe mirrors. Use "raidctl -S raid1"
while the first mirror is rewriting the parity.

>Fix:
None provided.

>Release-Note:
>Audit-Trail:

From: "Greg A. Woods" <woods@weird.com>
To: tron@zhadum.de
Cc: NetBSD GNATS submissions and followups <gnats-bugs@gnats.netbsd.org>,
  NetBSD Bugs and PR posting List <netbsd-bugs@NetBSD.ORG>
Subject: Re: bin/26741: "raidctl -S" doesn't always work as expected
Date: Tue, 24 Aug 2004 17:08:04 -0400 (EDT)

 [ On Monday, August 23, 2004 at 14:23:22 (+0200), Matthias Scheler wrote: ]
 > Subject: bin/26741: "raidctl -S" doesn't always work as expected
 >

 I think this wee patch may help make it do what you want:

 Index: sbin/raidctl/raidctl.c
 ===================================================================
 RCS file: /cvs/master/m-NetBSD/main/src/sbin/raidctl/raidctl.c,v
 retrieving revision 1.30
 diff -u -r1.30 raidctl.c
 --- sbin/raidctl/raidctl.c	21 Mar 2002 00:27:11 -0000	1.30
 +++ sbin/raidctl/raidctl.c	11 Jul 2004 17:40:38 -0000
 @@ -301,7 +301,7 @@
  		}
  		break;
  	case RAIDFRAME_CHECK_RECON_STATUS_EXT:
 -		check_status(fd,1);
 +		check_status(fd,verbose);
  		break;
  	case RAIDFRAME_GET_INFO:
  		if (do_output)
 @@ -860,10 +860,14 @@
  	int fd;
  	int meter;
  {
 +	int is_clean = 0;
  	int recon_percent_done = 0;
  	int parity_percent_done = 0;
  	int copyback_percent_done = 0;

 +	do_ioctl(fd, RAIDFRAME_CHECK_PARITY, &is_clean,
 +		 "RAIDFRAME_CHECK_PARITY");
 +	printf("Parity status: %s\n", is_clean ? "clean" : "DIRTY");
  	do_ioctl(fd, RAIDFRAME_CHECK_RECON_STATUS, &recon_percent_done, 
  		 "RAIDFRAME_CHECK_RECON_STATUS");
  	printf("Reconstruction is %d%% complete.\n", recon_percent_done);


 -- 
 						Greg A. Woods

 +1 416 218-0098                  VE3TCP            RoboHack <woods@robohack.ca>
 Planix, Inc. <woods@planix.com>          Secrets of the Weird <woods@weird.com>

From: Matthias Scheler <tron@zhadum.de>
To: NetBSD Bugs and PR posting List <netbsd-bugs@NetBSD.ORG>
Cc: NetBSD GNATS submissions and followups <gnats-bugs@gnats.netbsd.org>
Subject: Re: bin/26741: "raidctl -S" doesn't always work as expected
Date: Wed, 25 Aug 2004 08:32:29 +0200

 On Tue, Aug 24, 2004 at 05:08:04PM -0400, Greg A. Woods wrote:
 > I think this wee patch may help make it do what you want:
 > 
 > Index: sbin/raidctl/raidctl.c
 > ===================================================================
 > RCS file: /cvs/master/m-NetBSD/main/src/sbin/raidctl/raidctl.c,v
 > retrieving revision 1.30
 > diff -u -r1.30 raidctl.c
 > --- sbin/raidctl/raidctl.c	21 Mar 2002 00:27:11 -0000	1.30
 > +++ sbin/raidctl/raidctl.c	11 Jul 2004 17:40:38 -0000
 > @@ -301,7 +301,7 @@
 >  		}
 >  		break;
 >  	case RAIDFRAME_CHECK_RECON_STATUS_EXT:
 > -		check_status(fd,1);
 > +		check_status(fd,verbose);
 >  		break;
 >  	case RAIDFRAME_GET_INFO:
 >  		if (do_output)
 > @@ -860,10 +860,14 @@
 >  	int fd;
 >  	int meter;
 >  {
 > +	int is_clean = 0;
 >  	int recon_percent_done = 0;
 >  	int parity_percent_done = 0;
 >  	int copyback_percent_done = 0;
 >  
 > +	do_ioctl(fd, RAIDFRAME_CHECK_PARITY, &is_clean,
 > +		 "RAIDFRAME_CHECK_PARITY");
 > +	printf("Parity status: %s\n", is_clean ? "clean" : "DIRTY");
 >  	do_ioctl(fd, RAIDFRAME_CHECK_RECON_STATUS, &recon_percent_done, 
 >  		 "RAIDFRAME_CHECK_RECON_STATUS");
 >  	printf("Reconstruction is %d%% complete.\n", recon_percent_done);

 That would result in something like ...

 Parity status: DIRTY
 Reconstruction is 100% complete.
 Parity Re-write is 100% complete.
 Copyback is 100% complete.

 ... which looks wrong, too. I wonder why RAIDFRAME_CHECK_RECON_STATUS
 return 100% if the reconstruction hasn't even started.

 	Kind regards

 -- 
 Matthias Scheler                                  http://scheler.de/~matthias/

From: Greg Oster <oster@cs.usask.ca>
To: Matthias Scheler <tron@zhadum.de>
Cc: NetBSD Bugs and PR posting List <netbsd-bugs@NetBSD.ORG>,
  NetBSD GNATS submissions and followups <gnats-bugs@gnats.netbsd.org>
Subject: Re: bin/26741: "raidctl -S" doesn't always work as expected 
Date: Wed, 25 Aug 2004 07:44:38 -0600

 Matthias Scheler writes:
 > On Tue, Aug 24, 2004 at 05:08:04PM -0400, Greg A. Woods wrote:
 > > I think this wee patch may help make it do what you want:
 > > 
 > > Index: sbin/raidctl/raidctl.c
 > > ===================================================================
 > > RCS file: /cvs/master/m-NetBSD/main/src/sbin/raidctl/raidctl.c,v
 > > retrieving revision 1.30
 > > diff -u -r1.30 raidctl.c
 > > --- sbin/raidctl/raidctl.c	21 Mar 2002 00:27:11 -0000	1.30
 > > +++ sbin/raidctl/raidctl.c	11 Jul 2004 17:40:38 -0000
 > > @@ -301,7 +301,7 @@
 > >  		}
 > >  		break;
 > >  	case RAIDFRAME_CHECK_RECON_STATUS_EXT:
 > > -		check_status(fd,1);
 > > +		check_status(fd,verbose);
 > >  		break;
 > >  	case RAIDFRAME_GET_INFO:
 > >  		if (do_output)
 > > @@ -860,10 +860,14 @@
 > >  	int fd;
 > >  	int meter;
 > >  {
 > > +	int is_clean = 0;
 > >  	int recon_percent_done = 0;
 > >  	int parity_percent_done = 0;
 > >  	int copyback_percent_done = 0;
 > >  
 > > +	do_ioctl(fd, RAIDFRAME_CHECK_PARITY, &is_clean,
 > > +		 "RAIDFRAME_CHECK_PARITY");
 > > +	printf("Parity status: %s\n", is_clean ? "clean" : "DIRTY");
 > >  	do_ioctl(fd, RAIDFRAME_CHECK_RECON_STATUS, &recon_percent_done, 
 > >  		 "RAIDFRAME_CHECK_RECON_STATUS");
 > >  	printf("Reconstruction is %d%% complete.\n", recon_percent_done);
 > 
 > That would result in something like ...
 > 
 > Parity status: DIRTY
 > Reconstruction is 100% complete.
 > Parity Re-write is 100% complete.
 > Copyback is 100% complete.
 > 
 > ... which looks wrong, too. I wonder why RAIDFRAME_CHECK_RECON_STATUS
 > return 100% if the reconstruction hasn't even started.

 Cause it's somewhat bogus? :)  

 The "-S" option of raidctl is only meant to return the status of 
 parity rewriting, reconstruction, or copyback.  It's not meant to 
 tell you if the parity is dirty or not.  

 Arguably, "-S" should just go away, as the information it provides is 
 already in "-s" (where it is also too verbose, and shouldn't be 
 reporting 100% complete when it hasn't started anything.)  It's on my 
 "todo" list to fix these, but it hasn't made it onto the priority 
 radar.

 Later...

 Greg Oster



From: "Greg A. Woods" <woods@planix.com>
To: Matthias Scheler <tron@zhadum.de>
Cc: NetBSD Bugs and PR posting List <netbsd-bugs@NetBSD.ORG>,
  NetBSD GNATS submissions and followups <gnats-bugs@gnats.netbsd.org>
Subject: Re: bin/26741: "raidctl -S" doesn't always work as expected
Date: Thu, 26 Aug 2004 19:31:03 -0400 (EDT)

 [ On Wednesday, August 25, 2004 at 08:32:29 (+0200), Matthias Scheler wrote: ]
 > Subject: Re: bin/26741: "raidctl -S" doesn't always work as expected
 >
 > That would result in something like ...
 > 
 > Parity status: DIRTY
 > Reconstruction is 100% complete.
 > Parity Re-write is 100% complete.
 > Copyback is 100% complete.

 Indeed it does.

 > ... which looks wrong, too.

 Well it certainly seems a heck of a lot better than saying "100%
 complete" about everything when there's an obvious problem with the RAID
 set....


 > I wonder why RAIDFRAME_CHECK_RECON_STATUS
 > return 100% if the reconstruction hasn't even started.

 It's the status of the _previous_ operation when one is not ongoing.

 That seems logical enough to me.  :-)

 How's it supposed to know what the operator might do in the future?

 -- 
 						Greg A. Woods
 						Planix, Inc.

 <woods@planix.com>     +1 416 489-5852 x122     http://www.planix.com/
Responsible-Changed-From-To: bin-bug-people->oster 
Responsible-Changed-By: oster 
Responsible-Changed-When: Fri Aug 27 15:57:37 UTC 2004 
Responsible-Changed-Why:  
I look after RAIDframe lossage. 
>Unformatted:

Home	PR Database Search