source: gpfs_3.1_ker2.6.20/lpp/mmfs/bin/mmchcluster @ 47

Last change on this file since 47 was 16, checked in by rock, 17 years ago
  • Property svn:executable set to *
File size: 36.9 KB
RevLine 
[16]1#!/bin/ksh
2# IBM_PROLOG_BEGIN_TAG
3# This is an automatically generated prolog.
4# 
5# 
6# 
7# Licensed Materials - Property of IBM
8# 
9# (C) COPYRIGHT International Business Machines Corp. 2000,2007
10# All Rights Reserved
11# 
12# US Government Users Restricted Rights - Use, duplication or
13# disclosure restricted by GSA ADP Schedule Contract with IBM Corp.
14# 
15# IBM_PROLOG_END_TAG
16# @(#)11 1.59.1.3 src/avs/fs/mmfs/ts/admin/mmchcluster.sh, mmfs, avs_rgpfs24, rgpfs24s009a 12/19/06 13:10:44
17###############################################################################
18#
19# Usage:
20#   mmchcluster {[-p PrimaryServer] [-s SecondaryServer]}
21# or
22#   mmchcluster -p LATEST
23# or
24#   mmchcluster {[-r RemoteShellCommand] [-R RemoteFileCopyCommand]}
25# or
26#   mmchcluster -C ClusterName
27# or
28#   mmchcluster -N {NodeDesc[,NodeDesc...] | NodeFile}
29#
30# where:
31#
32#   -p PrimaryServer    specifies the node to be used as the primary server
33#                       of the GPFS sdrfs data for this cluster.
34#
35#      LATEST           requests a check to be made that all currently
36#                       available nodes point to the correct primary and
37#                       backup server.
38#
39#   -s SecondaryServer  specifies the node to be used as the backup server
40#                       of the GPFS sdrfs data for this cluster (optional).
41#                       To remove a backup server, specify  -s "".
42#
43#   -r RemoteShellCommand   specifies the fully qualified pathname for
44#                       the remote shell program to be used by GPFS.
45#                       The default is /usr/bin/rsh.
46#
47#   -R RemoteFileCopyCommand  specifies the fully qualified pathname for
48#                       the remote file copy program to be used by GPFS.
49#                       The default is /usr/bin/rcp.
50#
51#   -C ClusterName      specifies a new name for the cluster.  If the name
52#                       contains dots it is assumed to be a fully qualified
53#                       domain name.  Otherwise, the domain will default
54#                       to the domain of the primary configuration server.
55#
56#   -N NodeDesc,NodeDesc,...  specifies a comma-separated list of node
57#                             descriptors that specify the admin node
58#                             interfaces to be used in the cluster.
59#                       The node descriptors have the format:
60#                         daemonNodeName:nodeRoles:adminNodeName:
61#                       The nodeRoles field is currently just a place-holder
62#                       and is ignored.
63#
64#   -N NodeFile         specifies a file of node descriptors that specify
65#                       the admin node interfaces to be used in the cluster.
66#                       The lines in the input file have the format:
67#                         daemonNodeName:nodeRoles:adminNodeName:
68#                       The nodeRoles field is currently just a place-holder
69#                       and is ignored.
70#
71# Note:  When used with the -p or -s options, this command will most
72#        likely be needed when the current primary server is not available
73#        and it will be impossible to obtain the sdr lock and protect
74#        against concurrent execution of some other mm command.
75#        Under such conditions, the user must assure that no other mm
76#        command is run until the completion of the mmchcluster command
77#        and that as many of the remaining nodes as possible are available.
78#
79###############################################################################
80
81# Include global declarations and service routines.
82. /usr/lpp/mmfs/bin/mmglobfuncs
83. /usr/lpp/mmfs/bin/mmsdrfsdef
84
85sourceFile="mmchcluster.sh"
86[[ -n $DEBUG || -n $DEBUGmmchcluster ]] && set -x
87$mmTRACE_ENTER "$*"
88
89
90# Local work files.  Names should be of the form:
91#   fn=${tmpDir}fn.${mmcmd}.$$
92allNodes=${tmpDir}allNodes.${mmcmd}.$$
93clientNodes=${tmpDir}clientNodes.${mmcmd}.$$
94inputNodes=${tmpDir}inputNodes.${mmcmd}.$$
95processedNodes=${tmpDir}processedNodes.${mmcmd}.$$
96initErrors=${tmpDir}initErrors.${mmcmd}.$$
97# Note: Do not include initErrors in LOCAL_FILES yet; we'll do it later.
98
99LOCAL_FILES=" $allNodes $clientNodes $inputNodes $processedNodes "
100
101
102# Local declarations
103
104usageMsg=359
105newNodeNumbers=""
106backupServer=""
107rshPath=""
108rcpPath=""
109integer nodeCount
110integer n
111rc=0
112
113Cflag=""
114Nflag=""
115pflag=""
116rflag=""
117Rflag=""
118sflag=""
119Carg=""
120parg=""
121rarg=""
122Rarg=""
123sarg=""
124otherOpt=""
125
126
127# Local functions
128
129
130##########################################################################
131#
132# Function:  Specify the admin network for the GPFS cluster.
133#
134# Input:    $1 - file or list of node descriptors containing the
135#                adapter information as follows:
136#                  daemonNodeName:nodeRoles:adminNodeName:
137#
138# Returns:   0 - no errors encountered
139#            non-zero - unexpected error
140#
141##########################################################################
142function specifyAdminNetwork  # <networkInfo>
143{
144  typeset sourceFile="mmchcluster.sh"
145  [[ -n $DEBUG || -n $DEBUGspecifyAdminNetwork ]] && set -x
146  $mmTRACE_ENTER "$*"
147  typeset networkInfo="$1"
148
149  typeset failedNodes sdrfsLine mmcommonOutput
150  typeset nodeLine nodeName nodeName2 nodeStatus
151# typeset nodeRoles
152  typeset hostResult nodeNumber adminNodeName adminIpa
153  typeset nodeError newPrimaryName newBackupName commitOptions
154
155  typeset rc=0
156  typeset changeMade=""
157  typeset fatalError=""
158  typeset sharedSdrservPort=""
159
160  # The input parameter may be either a list or a file.  Which is it?
161  if [[ -f $networkInfo ]]
162  then
163    # It is a file; verify its existence and create our own copy.
164    checkUserFile $networkInfo $inputNodes
165    [[ $? -ne 0 ]] && cleanupAndExit
166  else
167    # It is not a file, so it must be a list.
168    # Convert the input node list into a file.
169    $rm -f $inputNodes
170    IFS=','
171    for nodeDesc in $networkInfo
172    do
173      print -- "$nodeDesc" >> $inputNodes
174      checkForErrors "writing to $inputNodes" $?
175    done
176    IFS="$IFS_sv"    # Restore the default IFS setting.
177  fi
178
179  # Check the input data for correctness.
180  # We check all the records rather than stop on the first error.
181  $rm -f $processedNodes
182  $touch $processedNodes     # Ensure the tmp file exists even if empty.
183  IFS=":"                    # Change the field separator to ':'.
184  exec 3<&-
185  exec 3< $inputNodes
186  while read -u3 nodeLine
187  do
188    # Parse the line.
189    set -f ; set -- $nodeLine ; set +f
190    nodeName=$1
191#   nodeRoles=$2
192    nodeName2=$3
193    IFS="$IFS_sv"    # Restore the default IFS setting.
194
195    # Make sure neither node name is specified more than once.
196    $grep -qw $nodeName $processedNodes > /dev/null 2>&1
197    if [[ $? -eq 0 ]]
198    then
199      # The node name is specified twice.
200      printErrorMsg 347 $mmcmd $nodeName
201      fatalError=yes
202    fi
203
204    # Check the admin node name if it was specified.
205    if [[ -n $nodeName2 && $nodeName2 != $nodeName ]]
206    then
207      $grep -qw $nodeName2 $processedNodes > /dev/null 2>&1
208      if [[ $? -eq 0 ]]
209      then
210        # The node is specified twice.
211        printErrorMsg 347 $mmcmd $nodeName2
212        fatalError=yes
213      fi
214    fi  # end of if [[ -n $nodeName2 && $nodeName2 != $nodeName ]]
215
216    # Add the node names to the list of processed nodes.
217    print -- "${nodeName}:${nodeName2}" >> $processedNodes
218    checkForErrors "Writing to file $processedNodes" $?
219
220    IFS=":"  # Change the separator back to ":" for the next iteration.
221
222  done  # end of while read -u3 nodeLine
223
224  IFS="$IFS_sv"  # Restore the default IFS settings.
225
226  # Return to the caller if we encountered an error.
227  [[ -n $fatalError ]] && return 1
228
229  # Ensure that the local copy of the mmsdrfs is up-to-date.
230  # Set up trap exception handling and obtain the lock.
231  trap pretrap HUP INT QUIT KILL
232  gpfsInitOutput=$(gpfsInit $lockId)
233  setGlobalVar $? $gpfsInitOutput
234
235  # Stop here if the admin network support has not been activated yet.
236  if [[ $sdrfsFormatLevel -eq 0 ]]
237  then
238    print -u2 "$mmcmd:  The separate administration network support has not been enabled yet."
239    print -u2 "    Run \"mmchconfig release=LATEST\" to activate the new function."
240    cleanupAndExit
241  fi
242
243  # Determine the lookup order for resolving host names.
244  [[ $osName != AIX ]] && resolveOrder=$(setHostResolveOrder)
245
246  # Go through the current mmsdrfs file.  Increment the generation
247  # number and build the node name list that will be needed later.
248  # Remove all admin network related information.
249  $rm -f $newsdrfs $nodefile
250  newPrimaryName=""
251  newBackupName=""
252  IFS=":"                    # Change the field separator to ':'.
253  exec 3<&-
254  exec 3< $mmsdrfsFile
255  while read -u3 sdrfsLine
256  do
257    # Parse the line.
258    set -f ; set -A v -- - $sdrfsLine ; set +f
259
260    IFS="$IFS_sv"      # Restore the default IFS settings.
261    printLine=true     # Assume the line will be printed.
262
263    case ${v[$LINE_TYPE_Field]} in
264
265      $VERSION_LINE )  # This is the global header line.
266        # Save the version line for updating later.
267        versionLine=$(print_newLine)
268        printLine=false
269        ;;
270
271      $NODESET_HDR )
272        # If the daemon and the mmsdrserv tcp ports are shared,
273        # it will be necessary to ensure that the daemon is down
274        # on the config server nodes if there names will change.
275        if [[ -z ${v[$GETOBJECT_PORT_Field]} ||
276              ${v[$TCP_PORT_Field]} = ${v[$GETOBJECT_PORT_Field]} ]]
277        then
278          sharedSdrservPort=yes
279        fi
280        ;;
281
282      $MEMBER_NODE )   # This line describes a node.
283        # Add the reliable node name to nodefile.
284        print -- "${v[$REL_HOSTNAME_Field]}" >> $nodefile
285        checkForErrors "writing to file $nodefile" $?
286
287        # Reset the node error flag.
288        nodeError=""
289
290        # Obtain the data for this node from the node file.
291        nodeLine=$($awk -F: '                        \
292          $1 == "'${v[$DAEMON_NODENAME_Field]}'" ||  \
293          $1 == "'${v[$REL_HOSTNAME_Field]}'"    ||  \
294          $1 == "'${v[$NODE_NAME_Field]}'"       ||  \
295          $1 == "'${v[$ADMIN_SHORTNAME_Field]}'" ||  \
296          $1 == "'${v[$NODE_NUMBER_Field]}'"     ||  \
297          $1 == "'${v[$IPA_Field]}'" {               \
298            { print $0 }                             \
299            { exit }                                 \
300          }                                          \
301        ' $inputNodes)
302
303        if [[ -n $nodeLine ]]
304        then
305          # We found data for this node.  Parse the input.
306          IFS=":"              # Change the field separator to ':'.
307          set -f ; set -- $nodeLine ; set +f
308          nodeName=$1
309          nodeName2=$3
310          IFS="$IFS_sv"        # Restore the default IFS setting.
311
312          # Determine the daemon node name.
313          if [[ -n ${v[$DAEMON_NODENAME_Field]} ]]
314          then
315            daemonNodeName=${v[$DAEMON_NODENAME_Field]}
316          else
317            daemonNodeName=${v[$REL_HOSTNAME_Field]}
318          fi
319
320          # Did the user reset or specify the admin node name?
321          if [[ -z $nodeName2 ]]
322          then
323            # The admin node name was null, indicating "reset";
324            # set the admin node name to the daemon node name value.
325            adminNodeName=$daemonNodeName
326            adminShortName=${v[$NODE_NAME_Field]}
327
328          else
329            # The admin node name was not null, indicating "specify";
330            # Determine the IP address for the specified admin node name.
331            hostResult=$($host $nodeName2)
332            set -f ; set -- $hostResult ; set +f
333            adminNodeName=$1
334            adminShortName=${1%% *|.*} # Exclude everything after the first dot.
335            adminIpa=${3%%,*}
336
337            # Check that the admin node name has a valid IP address.
338            if [[ -z $adminIpa ]]
339            then
340              # An invalid node name was specified.
341              printErrorMsg 54 $mmcmd $nodeName2
342              fatalError=yes
343              break
344            fi
345
346            # Invoke the checkAdapter function to ensure that
347            # the specified adapter interface exists on the node.
348            mmcommonOutput=$($mmcommon on1 ${v[$REL_HOSTNAME_Field]}  \
349               checkAdapter $adminIpa 2> $errMsg)
350            rc=$?
351            set -f ; set -- $mmcommonOutput ; set +f
352            nodeStatus=$1
353            if [[ $rc != 0 || $nodeStatus != success ]]
354            then
355              # The checkAdapter call failed.
356              # We will not define a new admin node name for this node
357              # but we will continue to process the remaining nodes.
358              # Tell the world what went wrong with this node.
359              if [[ $nodeStatus = ipa_alias ]]
360              then
361                # IP address aliasing is not supported.
362                printErrorMsg 476 $mmcmd $nodeName2
363              elif [[ $nodeStatus = ipa_missing ]]
364              then
365                # The admin IP address is not known on the node.
366                printErrorMsg 154 $mmcmd $nodeName2 ${v[$REL_HOSTNAME_Field]}
367              elif [[ $rc = $MM_HostDown || $rc = $MM_ConnectTimeout ]]
368              then
369                # The node cannot be reached.
370                printErrorMsg 340 $mmcmd ${v[$REL_HOSTNAME_Field]}
371              else
372                # Unexpected error.  Display all possible error messages.
373                [[ -s $errMsg ]] && $cat $errMsg 1>&2
374                [[ $rc -eq 0 ]] && rc=1
375                checkForErrors "checkAdapter ${v[$REL_HOSTNAME_Field]}" $rc
376              fi
377
378              # Append the node name to the list of failed nodes and
379              # set a flag to indicate the node name did not check out.
380              failedNodes="${failedNodes}\n\t${nodeName}"
381              nodeError=yes
382
383            fi  # end of if [[ $rc != 0 || $nodeStatus != success ]]
384
385          fi  # end of if [[ -z $nodeName2 ]]
386
387          # Update the member line if there was no error.
388          if [[ -z $nodeError ]]
389          then
390            # Remember the new primary or backup server name for updating
391            # the version line later if this is one of those servers.
392            [[ ${v[$REL_HOSTNAME_Field]} = $primaryServer ]] &&  \
393              newPrimaryName=$adminNodeName
394            [[ ${v[$REL_HOSTNAME_Field]} = $backupServer ]]  &&  \
395              newBackupName=$adminNodeName
396
397            # Things checked out ok.  Set the node name fields.
398            v[$DAEMON_NODENAME_Field]=$daemonNodeName
399            v[$REL_HOSTNAME_Field]=$adminNodeName
400            v[$ADMIN_SHORTNAME_Field]=$adminShortName
401            changeMade=yes
402          fi
403
404          $rm -f $errMsg
405
406        fi  # end of if [[ -n $nodeLine ]]
407        ;;
408
409      * )  # We are not interested in any other lines.
410        ;;
411
412    esac  # end of case ${v[$LINE_TYPE_Field]} in
413
414    # Unless suppressed, write the line to the new mmsdrfs file.
415    if [[ $printLine = true ]]
416    then
417      print_newLine >> $newsdrfs
418      checkForErrors "writing to file $newsdrfs" $?
419    fi
420
421    IFS=":"  # Change the separator back to ":" for the next iteration.
422
423  done  # end of while read -u3
424
425  IFS="$IFS_sv"  # Restore the default IFS settings.
426
427  # Go through the mmsdrfs file to update the NSD servers admin node names.
428  $rm -f $tmpsdrfs
429  IFS=":"
430  exec 3<&-
431  exec 3< $newsdrfs
432  while read -u3 sdrfsLine
433  do
434    # Parse the line.
435    set -f ; set -A v -- - $sdrfsLine ; set +f
436    IFS="$IFS_sv"
437
438    # Change some of the fields depending on the type of line.
439    case ${v[$LINE_TYPE_Field]} in
440
441      $SG_DISKS )  # This is the line for some disk.
442
443        # If this disk is an NSD with a valid PVID value,
444        # make sure the daemon nsd server names are recorded.
445        if [[ ${v[$DISK_TYPE_Field]} = nsd && -n ${v[$PVID_Field]} ]]
446        then
447          # If a server node was specified, check that it is valid and
448          # convert it to get the potentially new admin adapter name.
449          # We determine whether a server was specified by checking for an
450          # admin nsd server name, but we do not use that name for finding
451          # the node information, since the old admin node name may
452          # no longer exist as a result of the update we just did.
453          # We use the daemon node name to find the node instead,
454          # since mmchcluster -N does not change daemon node names.
455          if [[ -n ${v[$NSD_PRIMARY_NODE_Field]} ]]
456          then
457            # If no daemon node name has yet been recorded for the
458            # primary NSD server, determine and store it now.
459            server=${v[$DAEMON_NSD_PRIMARY_Field]}
460            if [[ -z $server ]]
461            then
462              server=$(checkAndConvertNodeValue  \
463                 ${v[$NSD_PRIMARY_NODE_Field]} $DAEMON_NODENAME_Field)
464              checkForErrors "checkAndConvertNodeValue" $?
465              v[$DAEMON_NSD_PRIMARY_Field]=$server
466            fi
467            # Use the primary server's daemon node name to obtain
468            # the primary server's admin node name.
469            v[$NSD_PRIMARY_NODE_Field]=$(checkAndConvertNodeValue  \
470               $server $REL_HOSTNAME_Field $newsdrfs)
471            checkForErrors "checkAndConvertNodeValue $server" $?
472          fi
473          if [[ -n ${v[$NSD_BACKUP_NODE_Field]} ]]
474          then
475            # If no daemon node name has yet been recorded for the
476            # backup NSD server, determine and store it now.
477            backup=${v[$DAEMON_NSD_BACKUP_Field]}
478            if [[ -z $backup ]]
479            then
480              backup=$(checkAndConvertNodeValue  \
481                 ${v[$NSD_BACKUP_NODE_Field]} $DAEMON_NODENAME_Field)
482              checkForErrors "checkAndConvertNodeValue" $?
483              v[$DAEMON_NSD_BACKUP_Field]=$backup
484            fi
485            # Use the backup server's daemon node name to obtain
486            # the backup server's admin node name.
487            v[$NSD_BACKUP_NODE_Field]=$(checkAndConvertNodeValue  \
488               $backup $REL_HOSTNAME_Field $newsdrfs)
489            checkForErrors "checkAndConvertNodeValue $backup" $?
490          fi
491        fi  # end of if (v[$DISK_TYPE_Field] == "nsd" && -n v[$PVID_Field])
492        ;;
493
494      * )  # We are not interested in any other lines.
495        ;;
496
497    esac  # end Change some of the fields
498
499    # Build and write the line to the temp version of the mmsdrfs file.
500    print_newLine >> $tmpsdrfs
501    checkForErrors "writing to file $tmpsdrfs" $?
502
503    IFS=":"  # Change the separator back to ":" for the next iteration.
504
505  done  # end while read -u3 sdrfsLine
506
507  IFS="$IFS_sv"  # Restore the default IFS settings.
508
509  # If a fatal error occurred, or if no changes were made,
510  # release the lock, report any failed nodes, and return.
511  if [[ -n $fatalError || -z $changeMade ]]
512  then
513    freeLockOnServer $primaryServer $ourNodeNumber >/dev/null
514    if [[ -n $failedNodes ]]
515    then
516      # Administrative node names were not defined for nodes ...
517      printErrorMsg 174 $mmcmd $failedNodes
518    fi
519    if [[ -n $fatalError ]]
520    then
521      printErrorMsg 389 $mmcmd         # The command failed.
522    else
523      printErrorMsg 387 $mmcmd $mmcmd  # Command quitting due to no valid nodes.
524    fi
525    return 1
526  fi
527
528  # Create the updated version line and add it to the new sdrfs file.
529  # The generation number is incremented and the server names may change.
530  IFS=":"                    # Change the field separator to ':'.
531  set -f ; set -A v -- - $versionLine ; set +f
532  IFS="$IFS_sv"              # Restore the default IFS setting.
533  newGenNumber=${v[$SDRFS_GENNUM_Field]}+1
534  v[$SDRFS_GENNUM_Field]=$newGenNumber
535  [[ -n $newPrimaryName ]] && v[$PRIMARY_SERVER_Field]=$newPrimaryName
536  [[ -n $newBackupName ]]  && v[$BACKUP_SERVER_Field]=$newBackupName
537  print_newLine >> $tmpsdrfs
538  checkForErrors "writing to file $tmpsdrfs" $?
539
540  # If the GPFS and mmsdrserv daemons share the same tcp port number,
541  # and the names of the primary or backup configuration servers are
542  # changing, it is necessary to ensure that the GPFS daemon is down
543  # on the server nodes and the mmsdrserv daemon is restarted.
544  # Otherwise, the server nodes will continue giving (stale) Gpfs object
545  # or return ESDR_NOT_SERVER errors.
546  if [[ -n $sharedSdrservPort && ( -n $newPrimaryName || -n $newBackupName ) ]]
547  then
548    # Get the names of the config servers.
549    print -- "${v[$PRIMARY_SERVER_Field]}\n${v[$BACKUP_SERVER_Field]}" > $tmpNodes
550    checkForErrors "writing to file $tmpNodes" $?
551
552    # Verify the daemon is down; do not lock the Gpfs object.
553    printInfoMsg 453
554    verifyDaemonInactive $tmpNodes
555    [[ $? -ne 0 ]] && return 1
556
557    commitOptions="initLocalNodeData,KILLSDRSERV"
558  else
559    commitOptions="initLocalNodeData"
560  fi  # end of if [[ -n $sharedSdrservPort ]]
561
562  # Make sure the new sdrfs file is properly sorted.
563  LC_ALL=C $SORT_MMSDRFS $tmpsdrfs -o $newsdrfs
564
565  # Put the new mmsdrfs file into the sdr.  This will make the newly-added
566  # admin nodes visible to the rest of the nodes in the cluster.
567  trap "" HUP INT QUIT KILL
568  gpfsObjectInfo=$(commitChanges $nsId $nsId  \
569     $gpfsObjectInfo $newGenNumber $newsdrfs $primaryServer $commitOptions)
570  rc=$?
571  if [[ $rc -ne 0 ]]
572  then
573    # We were unable to replace the file in the sdr.
574    printErrorMsg 381 $mmcmd
575    return 1
576  fi
577
578  # Unlock the sdr.
579  freeLockOnServer $primaryServer $ourNodeNumber >/dev/null
580  trap posttrap HUP INT QUIT KILL
581
582  # Propagate the new mmsdrfs file to all nodes in the cluster.
583  # This process is asynchronous.
584  propagateSdrfsFile async $nodefile $newsdrfs $newGenNumber initLocalNodeData
585
586  # Report any nodes that did not check successfully.
587  if [[ -n $failedNodes ]]
588  then
589    # Administrative node names were not defined for nodes ...
590    printErrorMsg 174 $mmcmd $failedNodes
591  fi
592
593  return 0
594
595}  #----- end of function specifyAdminNetwork -------------------
596
597
598###################################################################
599# This function is called if there is an interrupt after the new
600# mmsdrfs file was committed on the new primary and backup servers
601# but before the change was propagated to the rest of the nodes.
602###################################################################
603function localPosttrap
604{
605  $mmTRACE_ENTER "$*"
606
607  # Tell the guy which nodes must be up and which command to run.
608  printErrorMsg 350 $mmcmd "\n\t$newPrimaryServer\t$newBackupServer"
609  printErrorMsg 344 $mmcmd "mmchcluster -p LATEST"
610  cleanupAndExit 2
611
612}  #----- end of function localPosttrap ------------------------
613
614
615
616######################
617# Mainline processing
618######################
619
620
621###################################################
622# Process the command arguments.
623###################################################
624[[ $arg1 = '-?' || $arg1 = '-h' || $arg1 = '--help' || $arg1 = '--' ]] && \
625  syntaxError "help" $usageMsg
626
627[[ $argc -lt 2  ]] && \
628  syntaxError "missingArgs" $usageMsg
629
630while getopts :C:N:p:r:R:s: OPT
631do
632  case $OPT in
633
634    C) # cluster name
635       [[ -n $Cflag ]] && syntaxError "multiple" $noUsageMsg "-$OPT"
636       Cflag="-$OPT"
637       Carg=$OPTARG
638       ;;
639
640    N) # define/replace secondary network
641       [[ -n $Nflag ]] && syntaxError "multiple" $noUsageMsg "-$OPT"
642       Nflag="-$OPT"
643       Narg=$OPTARG
644       ;;
645
646    p) # primary server
647       [[ -n $pflag ]] && syntaxError "multiple" $noUsageMsg "-$OPT"
648       pflag="-$OPT"
649       parg=$OPTARG
650       otherOpt="-$OPT"
651       ;;
652
653    r) # remote shell command
654       [[ -n $rflag ]] && syntaxError "multiple" $noUsageMsg "-$OPT"
655       rflag="-$OPT"
656       rarg=$OPTARG
657       [[ $rarg = ${rarg#/} ]] && \
658         syntaxError "absolutePath_2" $noUsageMsg "-$OPT" "$rarg"
659       otherOpt="-$OPT"
660       ;;
661
662    R) # remote file copy command
663       [[ -n $Rflag ]] && syntaxError "multiple" $noUsageMsg "-$OPT"
664       Rflag="-$OPT"
665       Rarg=$OPTARG
666       [[ $Rarg = ${Rarg#/} ]] && \
667         syntaxError "absolutePath_2" $noUsageMsg "-$OPT" "$Rarg"
668       otherOpt="-$OPT"
669       ;;
670
671    s) # secondary server
672       [[ -n $sflag ]] && syntaxError "multiple" $noUsageMsg "-$OPT"
673       sflag="-$OPT"
674       sarg=$OPTARG
675       otherOpt="-$OPT"
676       ;;
677
678    +[CNprRs]) # Invalid option
679       syntaxError "invalidOption" $usageMsg $OPT
680       ;;
681
682    :) # Missing argument
683       syntaxError "missingValue" $usageMsg $OPTARG
684       ;;
685
686    *) # Invalid option
687       syntaxError "invalidOption" $usageMsg $OPTARG
688       ;;
689  esac
690
691done
692
693shift OPTIND-1
694[[ $# != 0 ]] && syntaxError "extraArg" $usageMsg $1
695
696[[ -n $sflag && $parg = LATEST ]] && \
697  syntaxError "invalidCombination" $usageMsg "-s" "-p LATEST"
698
699[[ -n $rflag && -n $pflag ]] && \
700  syntaxError "invalidCombination" $usageMsg "-r" "-p"
701
702[[ -n $rflag && -n $sflag ]] && \
703  syntaxError "invalidCombination" $usageMsg "-r" "-s"
704
705[[ -n $Rflag && -n $pflag ]] && \
706  syntaxError "invalidCombination" $usageMsg "-R" "-p"
707
708[[ -n $Rflag && -n $sflag ]] && \
709  syntaxError "invalidCombination" $usageMsg "-R" "-s"
710
711# The primary GPFS cluster configuration server cannot be removed.
712[[ -n $pflag && $parg = "" ]] && \
713  syntaxError "missingValue" $usageMsg "-p"
714
715[[ -n $Nflag && -n $otherOpt ]] && \
716  syntaxError "invalidCombination"  $usageMsg "-N" "$otherOpt"
717
718[[ -n $Cflag && -n $otherOpt ]] && \
719  syntaxError "invalidCombination" $usageMsg "-C" "$otherOpt"
720
721
722#############################################################################
723# If the request is to change a remote command, invoke the mmsetrcmd script.
724# Keep in mind that rarg and Rarg may include options for the respective
725# commands and, therefore, must always be quoted.
726#############################################################################
727if [[ -n $rflag || -n $Rflag ]]
728then
729  if [[ -z $Rflag ]]
730  then
731    $mmsetrcmd "$rflag" "$rarg"
732    rc=$?
733  elif [[ -z $rflag ]]
734  then
735    $mmsetrcmd "$Rflag" "$Rarg"
736    rc=$?
737  else
738    $mmsetrcmd "$rflag" "$rarg" "$Rflag" "$Rarg"
739    rc=$?
740  fi
741  cleanupAndExit $rc
742fi
743
744
745#############################################################
746# If the request is to specify changes to the admin network,
747# invoke the function to do the work and exit.
748#############################################################
749if [[ -n $Nflag ]]
750then
751  specifyAdminNetwork "$Narg"
752  cleanupAndExit $?
753fi
754
755
756########################################################
757# If the request is to change the cluster name,
758# invoke the mmsetrcmd script.
759########################################################
760if [[ -n $Cflag ]]
761then
762  $mmsetrcmd "$Cflag" "$Carg"
763  cleanupAndExit $?
764fi
765
766
767#################################################################
768# Set up trap exception handling and call the gpfsInit function.
769# It will attempt to ensure that the local copy of the mmsdrfs
770# and the rest of the GPFS system files are up-to-date.
771# Try to get the lock but do not fail if this is not possible.
772#################################################################
773trap pretrap HUP INT QUIT KILL
774
775if [[ $parg = LATEST ]]
776then
777  # The LATEST keyword was specified.  Try to obtain the
778  # most recent mmsdrfs file (i.e., the mmsdrfs file with the
779  # highest gen number) among all the nodes in the cluster.
780  # To do that, use the local mmsdrfs file as a starting point.
781  getNodeList $REL_HOSTNAME_Field $HOME_CLUSTER $mmsdrfsFile > $allNodes
782  gpfsInitOutput=$(gpfsInitFromNonServer $allNodes $mmsdrfsFile)
783  rc=$?
784
785else
786  # The LATEST keyword was not specified.  Try to obtain
787  # the mmsdrfs file from one of the servers with locking.
788  gpfsInitOutput=$(gpfsInit $lockId 2> $initErrors)
789  rc=$?
790  LOCAL_FILES="$LOCAL_FILES $initErrors "
791  if [[ $rc -ne 0 ]]
792  then
793    # We failed to get the sdrfs file with a lock.  Check whether
794    # some other mm command currently holds the lock.  If yes, give up.
795    $grep -e "Timed out waiting for lock:  Try again later."  \
796          -e "6027-1229"  $initErrors > /dev/null 2>&1
797    ec=$?
798    if [[ $ec -eq 0 ]]
799    then
800      # Display the messages from gpfsInit.
801      $cat $initErrors | \
802         $grep -v -e "6027-1227" -e "file is locked. Retrying..." 1>&2
803      cleanupAndExit
804    fi
805
806    # We failed to get the sdrfs file with a lock.  Display any messages.
807    $cat $initErrors 1>&2
808    # Processing continues.
809    printErrorMsg 437 $mmcmd
810
811    # Now try the gpfsInit again, but this time do not ask for a lock.
812    # If there is a backup server, and if it is available,
813    # we should be able to get the latest GPFS system files from there.
814    gpfsInitOutput=$(gpfsInit nolock 2>/dev/null)
815    rc=$?
816    if [[ $rc -ne 0 ]]
817    then
818      # We also failed to get the sdrfs file without locking.  Now try
819      # to obtain the most recent mmsdrfs file (i.e., the mmsdrfs file
820      # with the highest gen number) among all the nodes in the cluster.
821      # To do that, use the local mmsdrfs file as a starting point.
822      getNodeList $REL_HOSTNAME_Field $HOME_CLUSTER $mmsdrfsFile > $allNodes
823      gpfsInitOutput=$(gpfsInitFromNonServer $allNodes $mmsdrfsFile)
824      rc=$?
825    fi
826  fi
827fi   # end of if [[ $parg = LATEST ]]
828
829# Check whether we succeeded in obtaining the desired mmsdrfs file.
830if [[ $rc -ne 0 ]]
831then
832  # Not enough nodes are available.
833  printErrorMsg 378 $mmcmd
834  cleanupAndExit
835fi
836
837# Parse the output from the init function.
838setGlobalVar $rc $gpfsInitOutput
839
840if [[ $MMMODE = single ]]
841then
842  # Command currently not valid for cluster type single.
843  printErrorMsg 376 $mmcmd single
844  cleanupAndExit
845fi
846
847if [[ $MMMODE != lc ]]
848then
849  # Unknown GPFS nodeset type
850  printErrorMsg 338 $mmcmd $MMMODE
851  cleanupAndExit
852fi
853
854
855#######################################################
856# Determine the reliable hostnames of the new servers.
857#######################################################
858if [[ -n $pflag && $parg != LATEST ]]
859then
860  # Find the name of the primary server.
861  newPrimaryServer=$(checkAndConvertNodeValue $parg $REL_HOSTNAME_Field)
862  if [[ $? -ne 0 ]]
863  then
864    printErrorMsg 352 $mmcmd $parg
865    cleanupAndExit
866  fi
867else
868  # If -p not specified, the primary server remains the same.
869  newPrimaryServer=$primaryServer
870fi  # end of if [[ -n $parg && $parg != LATEST ]]
871
872if [[ -n $sflag ]]
873then
874  if [[ -n $sarg ]]
875  then
876    # Find the name of the secondary server.
877    newBackupServer=$(checkAndConvertNodeValue $sarg $REL_HOSTNAME_Field)
878    if [[ $? -ne 0 ]]
879    then
880      printErrorMsg 352 $mmcmd $sarg
881      cleanupAndExit
882    fi
883  else
884    # We are deleting the backup server (-s "" was specified).
885    newBackupServer=""
886  fi
887else
888  # If -s not specified, the backup server remains the same.
889  newBackupServer=$backupServer
890fi  # end of if [[ -n $sarg ]]
891
892# Cross check the two server names.
893if [[ $newBackupServer = $newPrimaryServer ]]
894then
895  # The same node was specified as primary and backup server.
896  printErrorMsg 346 $mmcmd
897  cleanupAndExit
898fi
899
900# Check whether anything needs to be done at all.
901[[ $newPrimaryServer = $primaryServer &&  \
902   $newBackupServer  = $backupServer  &&  \
903   $parg != LATEST ]] &&                  \
904  cleanupAndExit 0    # Servers are already as desired.
905
906
907#################################################################
908# Go through the current mmsdrfs file.  Increment the generation
909# number and change the server names.  Create a file with the
910# reliable hostnames of all nodes in the cluster.
911#################################################################
912$rm -f $newsdrfs $allNodes $clientNodes
913IFS=":"         # Change the field separator to ':'.
914exec 3<&-
915exec 3< $mmsdrfsFile
916while read -u3 sdrfsLine
917do
918  # Parse the line.
919  set -f ; set -A v -- - $sdrfsLine ; set +f
920  IFS="$IFS_sv"    # Restore the default IFS settings.
921
922  # Change some of the fields depending on the type of line.
923  case ${v[$LINE_TYPE_Field]} in
924
925    $VERSION_LINE )
926      # Increment the generation number.
927      newGenNumber=${v[$SDRFS_GENNUM_Field]}+1
928      v[$SDRFS_GENNUM_Field]=$newGenNumber
929      v[$PRIMARY_SERVER_Field]=$newPrimaryServer
930      v[$BACKUP_SERVER_Field]=$newBackupServer
931      ;;
932
933    $NODESET_HDR )
934      # If the daemon and the mmsdrserv tcp ports are shared,
935      # it will be necessary to ensure that the daemon is down
936      # on the old and new config server nodes.
937      if [[ -z ${v[$GETOBJECT_PORT_Field]} ||
938            ${v[$TCP_PORT_Field]} = ${v[$GETOBJECT_PORT_Field]} ]]
939      then
940        daemonMustBeDown=yes
941      fi
942      ;;
943
944    $MEMBER_NODE )
945      # If this is our node, save the reliable name.
946      [[ ${v[$NODE_NUMBER_Field]} = $ourNodeNumber ]] &&  \
947        ourNodeName=${v[$REL_HOSTNAME_Field]}
948
949      # All nodes will go in the allNodes file.
950      print -- "${v[$REL_HOSTNAME_Field]}" >> $allNodes
951      checkForErrors "writing to file $allNodes" $?
952
953      # The server nodes and the local node will
954      # not go in the clientNodes file.
955      if [[ ${v[$REL_HOSTNAME_Field]} != $newPrimaryServer &&
956            ${v[$REL_HOSTNAME_Field]} != $newBackupServer  &&
957            ${v[$REL_HOSTNAME_Field]} != $ourNodeName ]]
958      then
959        print -- "${v[$REL_HOSTNAME_Field]}" >> $clientNodes
960        checkForErrors "writing to file $clientNodes" $?
961      fi
962      ;;
963
964    * )  # Pass all other lines without change.
965      ;;
966
967  esac  # end Change some of the fields
968
969  # Build and write the line to the new mmsdrfs file.
970  print_newLine >> $newsdrfs
971  checkForErrors "writing to file $newsdrfs" $?
972
973  IFS=":"  # Change the separator back to ":" for the next iteration.
974
975done  # end of while read -u3 sdrfsLine
976
977IFS="$IFS_sv"  # Restore the default IFS settings.
978
979
980#######################################################################
981# If the GPFS and mmsdrserv daemons share the same tcp port number,
982# it is necessary to ensure that the GPFS daemon is down on the old
983# and new configuration server nodes.  Otherwise, the old server nodes
984# will continue giving (stale) Gpfs object information, while the new
985# servers will not be able to respond to requests because the GPFS
986# daemon cannot assume mmsdrserv duties if it is already running.
987#######################################################################
988if [[ -n $daemonMustBeDown && $parg != LATEST ]]
989then
990  # Put the old and new server names in a file.
991  print -- "$primaryServer\n$backupServer\n"  \
992           "$newPrimaryServer\n$newBackupServer" > $tmpNodes
993  checkForErrors "writing to file $tmpNodes" $?
994
995  # Eliminate duplicate names.
996  $sort -u $tmpNodes -o $tmpNodes
997  checkForErrors "sort $tmpNodes" $?
998
999  # Verify the daemon is down; do not lock the Gpfs object.
1000  printInfoMsg 453
1001  verifyDaemonInactive $tmpNodes
1002  [[ $? -ne 0 ]] && cleanupAndExit
1003fi  # end of if [[ -n $daemonMustBeDown ]]
1004
1005
1006######################################################
1007# First, put the new mmsdrfs file on the two servers.
1008# This must succeed no matter what.
1009######################################################
1010trap "" HUP INT QUIT KILL
1011gpfsObjectInfo=$(commitChanges  \
1012   $nsId $nsId $gpfsObjectInfo $newGenNumber $newsdrfs  \
1013   $newPrimaryServer FORCE $newBackupServer)
1014rc=$?
1015if [[ $rc -ne 0 ]]
1016then
1017  # Cannot replace file in the sdr.
1018  printErrorMsg 381 $mmcmd
1019
1020  # The mmchcluster failed - get out.
1021  # Tell the guy which nodes must be up and which command to run.
1022  printErrorMsg 350 $mmcmd "\n\t$newPrimaryServer\t$newBackupServer"
1023  printErrorMsg 344 $mmcmd "mmchcluster"
1024  cleanupAndExit
1025fi
1026
1027# Restore interrupts.
1028trap localPosttrap HUP INT QUIT KILL
1029
1030
1031#################################################
1032# Propagate the changes to the non-server nodes.
1033#################################################
1034if [[ $ourNodeName != $newPrimaryServer &&
1035      $ourNodeName != $newBackupServer ]]
1036then
1037  $cp $newsdrfs $mmsdrfsFile
1038  checkForErrors "writing to file $mmsdrfsFile" $?
1039fi
1040
1041if [[ -s $clientNodes ]]
1042then
1043  # Calculate the checksum of the new mmsdrfs file.
1044  sumOutput=$($sum $newsdrfs)
1045  checkForErrors "sum $newsdrfs" $?
1046  set -f ; set -- $sumOutput ; set +f
1047  newSum=$1
1048
1049#esjxx See if this can be replaced with pushSdr
1050  # Tell all client nodes to copy the file from us.
1051  $mmcommon onall $clientNodes $unreachedNodes copyRemoteFile  \
1052              $ourNodeName $mmsdrfsFile $mmsdrfsFile $newSum > $tmpfile 2>&1
1053  rc=$?
1054
1055  # Make a list of the nodes that were successfully updated.  For each
1056  # such node there will be a line in tmpfile that looks like this:
1057  #   nodename: copyRemoteFile:0
1058  updatedNodes=$($awk -F: ' {                         \
1059    if (($2 ~ "copyRemoteFile") && ($3 == "0")) {     \
1060      { print $1 }                                    \
1061    }                                                 \
1062  } ' $tmpfile)
1063  checkForErrors awk $?
1064
1065  # Determine the nodes that did not get the new data.
1066  exec 3<&-
1067  exec 3< $clientNodes
1068  while read -u3 nodeName
1069  do
1070    for goodNode in $updatedNodes
1071    do
1072      [[ $nodeName = $goodNode ]] &&  \
1073        break
1074    done
1075
1076    [[ $nodeName != $goodNode ]] &&  \
1077      failedNodes="${failedNodes}\n\t${nodeName}"
1078  done
1079
1080  # If any nodes failed, put out as much information as possible.
1081  if [[ -n $failedNodes ]]
1082  then
1083    # Collect error messages, if any, in file tmpfile2.
1084    $grep -v "copyRemoteFile:" $tmpfile > $tmpfile2
1085    [[ -s $tmpfile2 ]] &&  \
1086      $cat $tmpfile2 1>&2
1087
1088    # Tell the user which nodes failed.
1089    printErrorMsg 377 $mmcmd "$failedNodes"
1090    # Tell the guy which nodes must be up and which command to run.
1091    printErrorMsg 350 $mmcmd "\n\t$newPrimaryServer\t$newBackupServer"
1092    printErrorMsg 344 $mmcmd "mmchcluster -p LATEST"
1093    cleanupAndExit
1094  fi   # end if [[ -n $failedNodes ]]
1095
1096fi  # end if [[ ! -s $clientNodes ]]
1097
1098
1099##############################
1100# Unlock the sdr.
1101##############################
1102[[ $sdrLocked = yes ]] &&  \
1103  freeLockOnServer $primaryServer $ourNodeNumber > /dev/null
1104sdrLocked=no
1105trap posttrap HUP INT QUIT KILL
1106
1107# Issue "command was successful" message.
1108printErrorMsg 272 $mmcmd
1109cleanupAndExit 0
1110
Note: See TracBrowser for help on using the repository browser.