| Message ID | 20090911182308.GN15326@ether.msp.redhat.com (mailing list archive) |
|---|---|
| State | Changes Requested, archived |
| Delegated to: | Benjamin Marzinski |
| Headers | show |
Benjamin Marzinski wrote: > Instead of using the Target port's Asymmetric Access State to determine > priority, some users what to use the preferred port bit. This patch makes > a new prioritizer, based off the ALUA prioritizer, but using the preferred > port bit instead. > Please, no. This would mean that 'tpg_pref' would evaluate the PREF bit _only_, with total disregard to the other settings. We should be updating the existing 'alua' prioritizer to evaluate the PREF bit together with the existing AAS status; then we could be using it for all scenarios. I think I got a patch somewhere ... Cheers, Hannes
On Mon, Sep 14, 2009 at 08:24:39AM +0200, Hannes Reinecke wrote: > Benjamin Marzinski wrote: >> Instead of using the Target port's Asymmetric Access State to determine >> priority, some users what to use the preferred port bit. This patch makes >> a new prioritizer, based off the ALUA prioritizer, but using the preferred >> port bit instead. >> > Please, no. > > This would mean that 'tpg_pref' would evaluate the PREF bit _only_, > with total disregard to the other settings. > The PREF bit isn't static. It can change whenever the AAS status changes. I'd like to think that most devices do something sane like removing the PREF bit from offline and unavailable ports. Perhaps that is hoping for too much. So making sure that the preferred port is a sane choice makes sense. I'm not sure what you patch does, but I got the feeling that people wanted to choose, for instance, a preferred standby port over a non-preferred active port. If you have a patch that does that, I'm all for it. Otherwise, I can add some logic to make sure that the preferred port is a reasonable one (not unavailable or offline). How does that sound? -Ben > We should be updating the existing 'alua' prioritizer to evaluate > the PREF bit together with the existing AAS status; then we could > be using it for all scenarios. > > I think I got a patch somewhere ... > > Cheers, > > Hannes > -- > Dr. Hannes Reinecke zSeries & Storage > hare@suse.de +49 911 74053 688 > SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg > GF: Markus Rex, HRB 16746 (AG Nürnberg) > > -- > dm-devel mailing list > dm-devel@redhat.com > https://www.redhat.com/mailman/listinfo/dm-devel -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel
Index: multipath-tools-090902/libmultipath/prio.h =================================================================== --- multipath-tools-090902.orig/libmultipath/prio.h +++ multipath-tools-090902/libmultipath/prio.h @@ -16,6 +16,7 @@ * Known prioritizers for use in hwtable.c */ #define PRIO_ALUA "alua" +#define PRIO_TPG_PREF "tpg_pref" #define PRIO_CONST "const" #define PRIO_EMC "emc" #define PRIO_HDS "hds" Index: multipath-tools-090902/libmultipath/prioritizers/alua.c =================================================================== --- multipath-tools-090902.orig/libmultipath/prioritizers/alua.c +++ multipath-tools-090902/libmultipath/prioritizers/alua.c @@ -28,6 +28,12 @@ int get_alua_info(int fd) { +#ifdef TPG_PREF + char * aas_string[] = { + [AAS_OPTIMIZED] = "preferred", + [AAS_STANDBY] = "non-preferred", + }; +#else char * aas_string[] = { [AAS_OPTIMIZED] = "active/optimized", [AAS_NON_OPTIMIZED] = "active/non-optimized", @@ -35,6 +41,7 @@ get_alua_info(int fd) [AAS_UNAVAILABLE] = "unavailable", [AAS_TRANSITIONING] = "transitioning between states", }; +#endif int rc; int tpg; Index: multipath-tools-090902/libmultipath/prioritizers/alua_rtpg.c =================================================================== --- multipath-tools-090902.orig/libmultipath/prioritizers/alua_rtpg.c +++ multipath-tools-090902/libmultipath/prioritizers/alua_rtpg.c @@ -293,7 +293,11 @@ get_asymmetric_access_state(int fd, unsi "group.\n"); } else { PRINT_DEBUG("pref=%i\n", dscr->b0); +#ifdef TPG_PREF + rc = rtpg_tpg_dscr_get_pref(dscr); +#else rc = rtpg_tpg_dscr_get_aas(dscr); +#endif } } } Index: multipath-tools-090902/libmultipath/prioritizers/alua_spc3.h =================================================================== --- multipath-tools-090902.orig/libmultipath/prioritizers/alua_spc3.h +++ multipath-tools-090902/libmultipath/prioritizers/alua_spc3.h @@ -302,6 +302,12 @@ rtpg_tpg_dscr_get_aas(struct rtpg_tpg_ds return (d->b0 & 0x0f); } +static inline int +rtpg_tpg_dscr_get_pref(struct rtpg_tpg_dscr *d) +{ + return ((d->b0 & 0x80)?AAS_OPTIMIZED:AAS_STANDBY); +} + struct rtpg_data { unsigned char length[4]; /* size-4 */ struct rtpg_tpg_dscr data[0]; Index: multipath-tools-090902/libmultipath/prioritizers/Makefile =================================================================== --- multipath-tools-090902.orig/libmultipath/prioritizers/Makefile +++ multipath-tools-090902/libmultipath/prioritizers/Makefile @@ -11,6 +11,7 @@ LIBS = \ libprioemc.so \ libpriordac.so \ libprioalua.so \ + libpriotpg_pref.so \ libprionetapp.so \ libpriohds.so @@ -21,9 +22,18 @@ all: $(LIBS) libprioalua.so: alua.o alua_rtpg.o $(CC) $(SHARED_FLAGS) -o $@ $^ +libpriotpg_pref.so: pref.o pref_rtpg.o + $(CC) $(SHARED_FLAGS) -o $@ $^ + libprio%.so: %.o $(CC) $(SHARED_FLAGS) -o $@ $^ +pref.o: alua.c + $(CC) $(CFLAGS) -DTPG_PREF -c -o $@ $< + +pref_rtpg.o: alua_rtpg.c + $(CC) $(CFLAGS) -DTPG_PREF -c -o $@ $< + install: $(LIBS) $(INSTALL_PROGRAM) -m 755 libprio*.so $(DESTDIR)$(libdir)
Instead of using the Target port's Asymmetric Access State to determine priority, some users what to use the preferred port bit. This patch makes a new prioritizer, based off the ALUA prioritizer, but using the preferred port bit instead. Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com> --- libmultipath/prio.h | 1 + libmultipath/prioritizers/Makefile | 10 ++++++++++ libmultipath/prioritizers/alua.c | 7 +++++++ libmultipath/prioritizers/alua_rtpg.c | 4 ++++ libmultipath/prioritizers/alua_spc3.h | 6 ++++++ 5 files changed, 28 insertions(+) -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel