[patch]: [multipath-tool] Add check for removed device in rdac checker

[patch]: [multipath-tool] Add check for removed device in rdac checker

[Yanqing_Liu]

Hi,

This patch adds a check for removed device in rdac checker. After volume
access inquiry returns, the checker will check the peripheral qualifier
field to see if the device is deleted/unmapped. If so, mark path as
failed.

Signed off by: Yanqing Liu <Yanqing_Liu@Dell.com>

---

--- libcheckers/rdac.c.orig	2008-10-22 02:52:33.000000000 -0400
+++ libcheckers/rdac.c	2008-10-17 03:43:13.000000000 -0400
@@ -27,6 +27,7 @@
 #define MSG_RDAC_UP    "rdac checker reports path is up"
 #define MSG_RDAC_DOWN  "rdac checker reports path is down"
 #define MSG_RDAC_GHOST "rdac checker reports path is ghost"
+#define MSG_RDAC_DELETED "rdac checker reports path deleted"
 
 struct rdac_checker_context {
 	void * dummy;
@@ -90,7 +91,8 @@
 
 struct volume_access_inq
 {
-	char dontcare0[8];
+	char qualifier;
+	char dontcare0[7];
 	char avtcvp;
 	char dontcare1[39];
 };
@@ -104,6 +106,12 @@
 		MSG(c, MSG_RDAC_DOWN);
 		return PATH_DOWN;
 	}
+	
+	if (0 != inq.qualifier) {
+		MSG(c, MSG_RDAC_DELETED);
+		return PATH_DOWN;
+	}
+		
 
 	return ((inq.avtcvp & 0x1) ? PATH_UP : PATH_GHOST);  }

Re: [patch]: [multipath-tool] Add check for removed device in rdac checker

[Konrad Rzeszutek]

On Mon, Oct 27, 2008 at 10:11:28AM -0500, Yanqing_Liu@Dell.com wrote:
> Hi,
> 
> This patch adds a check for removed device in rdac checker. After volume
> access inquiry returns, the checker will check the peripheral qualifier
> field to see if the device is deleted/unmapped. If so, mark path as
> failed.

How does multipathd behaves when this fix is not in? Does
it still mark the path down? I would presume so, but it would take
longer (the default SCSI timeout value?)

> 
> Signed off by: Yanqing Liu <Yanqing_Liu@Dell.com>
> 
> ---
> 
> --- libcheckers/rdac.c.orig	2008-10-22 02:52:33.000000000 -0400
> +++ libcheckers/rdac.c	2008-10-17 03:43:13.000000000 -0400
> @@ -27,6 +27,7 @@
>  #define MSG_RDAC_UP    "rdac checker reports path is up"
>  #define MSG_RDAC_DOWN  "rdac checker reports path is down"
>  #define MSG_RDAC_GHOST "rdac checker reports path is ghost"
> +#define MSG_RDAC_DELETED "rdac checker reports path deleted"
>  
>  struct rdac_checker_context {
>  	void * dummy;
> @@ -90,7 +91,8 @@
>  
>  struct volume_access_inq
>  {
> -	char dontcare0[8];
> +	char qualifier;
> +	char dontcare0[7];
>  	char avtcvp;
>  	char dontcare1[39];
>  };
> @@ -104,6 +106,12 @@
>  		MSG(c, MSG_RDAC_DOWN);
>  		return PATH_DOWN;
>  	}
> +	
> +	if (0 != inq.qualifier) {
> +		MSG(c, MSG_RDAC_DELETED);
> +		return PATH_DOWN;
> +	}
> +		
>  
>  	return ((inq.avtcvp & 0x1) ? PATH_UP : PATH_GHOST);  }
> 
> 
> 
> 
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel

RE: [patch]: [multipath-tool] Add check for removeddevice in rdac checker

[Shyam_Iyer]

[-- Attachment #1.1: Type: text/plain, Size: 837 bytes --]

Konrad Wrote:
> On Mon, Oct 27, 2008 at 10:11:28AM -0500, Yanqing_Liu@Dell.com wrote:
> > Hi,
> >
> > This patch adds a check for removed device in rdac checker. After
> > volume access inquiry returns, the checker will check the peripheral
> > qualifier field to see if the device is deleted/unmapped. If so,
mark
> > path as failed.

> How does multipathd behaves when this fix is not in? Does it still
mark the path down? I would presume so, but it would take longer (the
default SCSI timeout value?)

Absolutely. But that would also be after some IOs on the failed path and
the device being marked offline. Instead if the controller device is
intelligent to support a feature to know if the path is down through an
inquiry then it is worth marking the path down earlier. 

Thanks for reviewing.

-Shyam Iyer.


[-- Attachment #1.2: Type: text/html, Size: 1249 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: [patch]: [multipath-tool] Add check for removeddevice in rdac checker

[Konrad Rzeszutek]

On Tue, Oct 28, 2008 at 03:11:08PM +0530, Shyam_Iyer@Dell.com wrote:
> Konrad Wrote:
> > On Mon, Oct 27, 2008 at 10:11:28AM -0500, Yanqing_Liu@Dell.com wrote:
> > > Hi,
> > >
> > > This patch adds a check for removed device in rdac checker. After
> > > volume access inquiry returns, the checker will check the peripheral
> > > qualifier field to see if the device is deleted/unmapped. If so,
> mark
> > > path as failed.
> 
> > How does multipathd behaves when this fix is not in? Does it still
> mark the path down? I would presume so, but it would take longer (the
> default SCSI timeout value?)
> 
> Absolutely. But that would also be after some IOs on the failed path and
> the device being marked offline. Instead if the controller device is
> intelligent to support a feature to know if the path is down through an
> inquiry then it is worth marking the path down earlier. 

I concur.
> 
> Thanks for reviewing.

You are welcome.

+1