[Orca-dev] RE[2]: Digest 336

[Rusty Carruth]

alegrand at wallace.com wrote:
> I am unclear about the hardware that Rusty is using  his short disk name
> is dad0.

Well, this is an Ultra 5 running Solaris 2.5.1.

I modified orcallator.se to tell me what the short names were as it added them,
and it got this list:

adding raw disk.  short name: dad0
adding raw disk.  short name: atapicd2
adding raw disk.  short name: fd0

(this is on SunOS msfree 5.5.1 Generic_103640-32 sun4u sparc SUNW,Ultra-5_10)

Over in the other building I found one machine that gave me this:

adding raw disk.  short name: sd0
adding raw disk.  short name: sd1
adding raw disk.  short name: sd2

(and that was on SunOS cass 5.5.1 Generic_103640-24 sun4u sparc SUNW,Ultra-1)

> Every short disk name that I have seen is "sdXX"  this maybe the source of
> the problem.  I am sure that I used this to recognize this disk.  So now
> I find myself it the same situation you keep having.  
> 
> I have an idea what the problem is  but I do not have the hardware it is 
> happening on to replicate it.
> 
> The best solution would be for Rusty and I figure out what needs to be
> changed.

Well, like I say, I did this (more comments follow the code stuff):

bash-2.04$ diff -u /tmp/orca*
--- /tmp/orcallator.se  Thu Oct  4 12:55:22 2001
+++ /tmp/orcallator.se.orig     Thu Oct  4 12:55:22 2001
@@ -470,7 +470,10 @@
   // already identified how many physical devices are present and they
   // are listed first we limit our self to mapping the first
   // RAW_disk_count device name.
+//printf("raw disk count is %d, changed to %d\n",RAW_disk_count,RAW_disk_count - 1);
+printf("raw disk count is %d\n",RAW_disk_count);
   for (i=0; i<RAW_disk_count; i++) {
+printf("I is now %d\n",i);
     strcpy(long_name,GLOBAL_disk_info[i].long_name);
     if (long_name[0] == 'c' && strchr(long_name, 's') == nil) {
       strcpy(short_name,GLOBAL_disk_info[i].short_name);
@@ -535,6 +538,8 @@
       strcpy(short_name, nkp[0].ks_name);
       if (short_name[0] != 'm' &&
           short_name[0] != 'n' &&
+         short_name[0] != 'f' && // fd fails on sun - rcarruth
+         short_name[1] != 'd' && // ditto - rcarruth
           strchr(short_name,',') == nil) {
         // Try to locate device.
         for (rdisk=0; rdisk<RAW_disk_count;rdisk++) {
@@ -550,6 +555,7 @@
           RAW_disk_map = 1;
           strcpy(RAW_disk[rdisk].long_name, short_name);
           strcpy(RAW_disk[rdisk].short_name, short_name);
+printf("adding raw disk.  short name: %s\n",short_name);
           RAW_disk[rdisk]._reads       = 0;
           RAW_disk[rdisk]._nread       = 0;
           RAW_disk[rdisk]._rlentime    = 0;
@@ -1041,31 +1047,48 @@
   }
 }
 
+/* state as a digit */
+int state_dig(int state) {
+  switch(state) {
+    case ST_WHITE: return '0'; /* OK states are zero */
+    case ST_BLUE:  return '0';
+    case ST_GREEN: return '0';
+    case ST_AMBER: return '1';/* bad states have a non-zero value.
+    case ST_RED:   return '2';
+    case ST_BLACK: return '4';
+    default: return '9';       /* invalid state */
+  }
+}
+
 measure_misc(long now, tm_t tm_now)
 {
   long   uptime;
   char   states[12];
   char   tm_buf[16];
+  char sum_states[24];
+  int statetmp;
 
   uptime = now - boot_time;
+  sum_states = "0 0 0 0 0 0 0 0 0 0 0";
   states = "wwwwwwwwwww";
   strftime(tm_buf, sizeof(tm_buf), "%T", tm_now);
-
-  states[0] = state_char(lr_disk$dr.state);
-  states[1] = state_char(lr_net$nr.state);
-  states[2] = state_char(lr_rpcclient$r.state);
-  states[3] = state_char(lr_swapspace$s.state);
-  states[4] = state_char(lr_ram$ram.state);
-  states[5] = state_char(lr_kmem$kmem.state);
-  states[6] = state_char(lr_cpu$cpu.state);
-  states[7] = state_char(lr_mutex$m.state);
-  states[8] = state_char(lr_dnlc$dnlc.state);
-  states[9] = state_char(lr_inode$inode.state);
-  states[10]= state_char(lr_tcp$tcp.state);
+  statetmp = lr_disk$dr.state;      states[0] = state_char(statetmp);   sum_states[0] = state_dig(statetmp);
+  statetmp = lr_net$nr.state;       states[1] = state_char(statetmp);    sum_states[2] = state_dig(statetmp);
+  statetmp = lr_rpcclient$r.state;  states[2] = state_char(statetmp);    sum_states[4] = state_dig(statetmp);
+  statetmp = lr_swapspace$s.state;  states[3] = state_char(statetmp);    sum_states[6] = state_dig(statetmp);
+  statetmp = lr_ram$ram.state;      states[4] = state_char(statetmp);    sum_states[8] = state_dig(statetmp);
+  statetmp = lr_kmem$kmem.state;    states[5] = state_char(statetmp);    sum_states[10] = state_dig(statetmp);
+  statetmp = lr_cpu$cpu.state;      states[6] = state_char(statetmp);    sum_states[12] = state_dig(statetmp);
+  statetmp = lr_mutex$m.state;      states[7] = state_char(statetmp);    sum_states[14] = state_dig(statetmp);
+  statetmp = lr_dnlc$dnlc.state;    states[8] = state_char(statetmp);    sum_states[16] = state_dig(statetmp);
+  statetmp = lr_inode$inode.state;  states[9] = state_char(statetmp);    sum_states[18] = state_dig(statetmp);
+  statetmp = lr_tcp$tcp.state;      states[10]= state_char(statetmp);    sum_states[20]= state_dig(statetmp);
 
   put_output(" timestamp",  sprintf("%10d", now));
   put_output("locltime",    tm_buf);
   put_output("DNnsrkcmdit", states);
+  put_output(" Dw Nw nw sw rw kw cw mw dw iw tw", sum_states);
+//             0  1  2  3  4  5  6  7  8  9  10
   put_output("  uptime",    sprintf("%8d", uptime));
 }
 
bash-2.04$ 


> Yes, I think this code does include the patch.

(yes, it does)

> So what do we really need here?  Can you send a patch to fix it?

I've got Version 1.28b4: Mar 27, 2001, with the above patches at this point.
The trick was here:

       if (short_name[0] != 'm' &&
           short_name[0] != 'n' &&
+         short_name[0] != 'f' && // fd fails on my sun ultra5 - rcarruth
+         short_name[1] != 'd' && // ditto - rcarruth

And it seems to work fine on all the machines I've tried it on (27 machines)

So, like everyone is saying - now what?  ;-)

rc

> alegrand at wallace.com wrote:
> > 
> > ]Blair,
> > 
> >         Does the version of orcallator  Rusty is using include the patch
> > I sent you early this year.   For the section that implements the rawdisk
> > code? (version 1.28b4)
> > 
> > I believe it should have stopped this problem permanently.  In any case
> > the problem is not a privillege issue.   The structure being searched
> > contains mappings between the c..... long disk name to  sdxxx short disk
> > names.   Somestimes the long disk name field is missing.  Sometimes the
> > list of mapppings also contains Disks that are nolonger in the system.
> > So you need to seach the table until you find everything you want.
> > Unfortunately
> > The length of as reported by GLOBAL_diskinfo_size  is also incorrect.
> > 
> > The line that accomplishes this is
> >    if (strchr(long_name, 's' != nil) { break;}
> > 
> > ALan
> >