Changeset a74ea78

Timestamp:

Jul 24, 2009 7:10:29 PM (15 years ago)

Author:

Ted Faber <faber@…>

Branches:

axis_example, compt_changes, info-ops, master, version-1.30, version-2.00, version-3.01, version-3.02

Children:

281c0ca

Parents:

bd3e314

Message:

Properly avoid terminating a starting experiment (this will have to get an
override) and be more conservative when starting ssh commands.

File:

• fedd/federation/experiment_control.py (modified) (5 diffs)

fedd/federation/experiment_control.py

@@ -530 +530 @@
         def ssh_cmd(self, user, host, cmd, wname=None, timeout=None):
             """
-            Run a remote command on host as user.  If debug is set, the action is
-            only logged.
+            Run a remote command on host as user.  If debug is set, the action
+            is only logged.  Commands are run without stdin, to avoid stray
+            SIGTTINs.
             """
-            sh_str = ("%s -o 'IdentitiesOnly yes' -o " + \
+            sh_str = ("%s -n -o 'IdentitiesOnly yes' -o " + \
                     "'StrictHostKeyChecking yes' -i %s %s@%s %s") % \
                     (self.ssh_exec, self.ssh_privkey_file, 
@@ -2033 +2034 @@
                     {'localname' : eid }, { 'fedid': copy.copy(expid) }
                 ],
-                'experimentStatus': 'started',
+                'experimentStatus': 'starting',
                 'experimentAccess': { 'X509' : expcert }
             }
@@ -2218 +2219 @@
                 rv['allocationLog'] = "".join(rv['log'])
                 del rv['log']
+            else:
+                rv['allocationLog'] = ""
 
             if rv['experimentStatus'] != 'active':
@@ -2266 +2269 @@
 
             # First make sure that the experiment creation is complete.
-            if fed_exp.has_key('experimentStatus'):
-                if fed_exp['experimentStatus'] == 'started':
+            status = fed_exp.get('experimentStatus', None)
+            if status:
+                if status == 'starting':
                     self.state_lock.release()
                     raise service_error(service_error.partial, 
@@ -2330 +2334 @@
             thread_pool.wait_for_all_done()
 
-            # release the allocations
-            for tb in tbparams.keys():
-                self.release_access(tb, tbparams[tb]['aid'])
+            # release the allocations (failed experiments have done this
+            # already)
+            if status != 'failed':
+                for tb in tbparams.keys():
+                    self.release_access(tb, tbparams[tb]['aid'])
 
             # Remove the terminated experiment