Do more clever locking to prevent two DCMs from stepping on each other.

Specifically:

	- Always check if the inprogress flag is set in the servers or
	  serverhosts tables before setting it.

	- When running dcm with specific services on the command line,
	  still do interval checking; if the DCM has been run too
	  recently, refuse to run.

	- Provide a -f/-force flag to override the interval checking.

diff --git a/dcm/dcm.pc b/dcm/dcm.pc
index 105ec650e3772a53b40ec1b085c30335c962de4e..2f583fe32ad47ca15d550ebb2b9ec3097e9853f5 100644 (file)
--- a/dcm/dcm.pc
+++ b/dcm/dcm.pc
@@ -38,19 +38,23 @@ void dbmserr(void);
 #define SQL_NO_MATCH 1403
 #define SOFT_FAIL(x) (((x) == MR_NO_MEM) || ((x) == MR_CANT_CONNECT) || ((x) == MR_CCONFIG) || ((x) == MR_DEADLOCK) || ((x) == MR_BUSY) || ((x) == MR_ABORT))
 
+/* argument parsing macro */
+#define argis(a, b) (!strcmp(*arg + 1, a) || !strcmp(*arg + 1, b))
+
 char whobuf[256], *whoami = whobuf, *db = "moira";
 
 enum { UNIQUE, DISTRIBUTED, REPLICATED };
 
 int main(int argc, char **argv)
 {
-  int i;
+  int i, force = 0;
   EXEC SQL BEGIN DECLARE SECTION;
   char buf[SERVERS_NAME_SIZE], *name;
   int enable;
   EXEC SQL END DECLARE SECTION;
   struct save_queue *sq;
   int status;
+  char **arg = argv;
 
   if (strchr(argv[0], '/'))
     strcpy(whoami, strrchr(argv[0], '/') + 1);
@@ -63,12 +67,28 @@ int main(int argc, char **argv)
   initialize_sms_error_table();
   initialize_krb_error_table();
 
+  while (++arg - argv < argc)
+    {
+      if (**arg == '-')
+       {
+         if (argis("f", "force"))
+           force++;
+         else
+           {
+             com_err(whoami, 0, "Usage: %s [-f] servicename", argv[0]);
+             exit(1);
+           }
+       }
+    }
+
   /* if services were specified on the command line, do just those ones */
   if (argc > 1)
     {
       for (i = 1; i < argc; i++)
        {
-         if (generate_service(argv[i], 1))
+         if (argv[i][0] == '-')
+           continue;
+         if (generate_service(argv[i], force))
            do_hosts(argv[i]);
        }
       exit(0);
@@ -113,7 +133,7 @@ int main(int argc, char **argv)
   /* Now run through list */
   while (sq_get_data(sq, &name))
     {
-      if (generate_service(name, 0))
+      if (generate_service(name, force))
        {
          switch (fork())
            {
@@ -142,7 +162,7 @@ int main(int argc, char **argv)
 int generate_service(char *name, int force)
 {
   EXEC SQL BEGIN DECLARE SECTION;
-  int interval, dfcheck, status;
+  int interval, dfcheck, status, inprogress;
   time_t now;
   const char *errmsg;
   EXEC SQL END DECLARE SECTION;
@@ -152,8 +172,8 @@ int generate_service(char *name, int force)
 
   EXEC SQL CONNECT :db IDENTIFIED BY :db;
 
-  EXEC SQL SELECT update_int, dfcheck INTO :interval, :dfcheck
-    FROM servers WHERE name = UPPER(:name);
+  EXEC SQL SELECT update_int, dfcheck, inprogress INTO :interval, :dfcheck,
+    :inprogress FROM servers WHERE name = UPPER(:name);
   if (sqlca.sqlcode == SQL_NO_MATCH)
     {
       com_err(whoami, 0, "No such service `%s'", name);
@@ -161,6 +181,17 @@ int generate_service(char *name, int force)
       return 0;
     }
 
+  /* Someone might try to run a DCM from the command line while the
+   * regular one is running, which will bypass the "interval" test.
+   * Check inprogress to make sure they don't stomp on themselves.
+   */
+  if (inprogress == 1)
+    {
+      com_err(whoami, 0, "DCM for service `%s' already in progress", name);
+      EXEC SQL COMMIT RELEASE;
+      return 0;
+    }
+
   time(&now);
 
   if ((interval * 60 + dfcheck < now) || force)
@@ -230,6 +261,12 @@ int generate_service(char *name, int force)
                         name, errmsg);
        }
     }
+  else
+    {
+      com_err(whoami, 0, "DCM for service `%s' has run too recently.", name);
+      com_err(whoami, 0, "Use the -force flag to force a DCM.");
+    }
+
   EXEC SQL COMMIT RELEASE;
   return 0;
 
@@ -246,7 +283,7 @@ void do_hosts(char *service)
   char server_type[SERVERS_TYPE_SIZE], host[MACHINE_NAME_SIZE], *name;
   char target[SERVERS_TARGET_FILE_SIZE], script[SERVERS_SCRIPT_SIZE];
   const char *errmsg;
-  int status = 0, dfgen, type, mid;
+  int status = 0, dfgen, type, mid, inprogress;
   time_t now;
   EXEC SQL END DECLARE SECTION;
   struct save_queue *sq;
@@ -256,8 +293,8 @@ void do_hosts(char *service)
 
   EXEC SQL CONNECT :db IDENTIFIED BY :db;
 
-  EXEC SQL SELECT dfgen, type, target_file, script
-    INTO :dfgen, :server_type, :target, :script
+  EXEC SQL SELECT dfgen, type, target_file, script, inprogress
+    INTO :dfgen, :server_type, :target, :script, :inprogress
     FROM servers WHERE name = UPPER(:service);
   if (!strncmp(strtrim(server_type), "REPLICAT", 8))
     type = REPLICATED;
@@ -268,6 +305,13 @@ void do_hosts(char *service)
   strtrim(target);
   strtrim(script);
 
+  if (inprogress == 1)
+    {
+      com_err(whoami, 0, "DCM for service `%s' already in progress", name);
+      EXEC SQL COMMIT RELEASE;
+      return;
+    }
+
   EXEC SQL DECLARE csr_hst1 CURSOR FOR
     SELECT m.name, m.mach_id FROM machine m, serverhosts sh
     WHERE sh.service = UPPER(:service)
@@ -290,6 +334,16 @@ void do_hosts(char *service)
   while (sq_get_data(sq, &name))
     {
       sq_get_data(sq, &mid);
+
+      EXEC SQL SELECT inprogress INTO :inprogress FROM serverhosts
+       WHERE service = UPPER(:service) AND mach_id = :mid;
+      if (inprogress == 1)
+       {
+         com_err(whoami, 0, "DCM for service `%s' to host `%s' already in progress", service, name);
+         EXEC SQL COMMIT RELEASE;
+         return;
+       }
+
       com_err(whoami, 0, "sending %s data to %s", service, name);
       EXEC SQL UPDATE serverhosts SET inprogress = 1
        WHERE service = UPPER(:service) AND mach_id = :mid;