Skip to content

Commit

Permalink
revert 94053a7
Browse files Browse the repository at this point in the history
  • Loading branch information
@mengxr
mengxr committed Nov 25, 2014
1 parent 4298ea5 commit f0b708b
Show file tree
Hide file tree
Showing 2 changed files with 27 additions and 60 deletions.
14 changes: 6 additions & 8 deletions docs/ec2-scripts.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,16 +12,14 @@ on the [Amazon Web Services site](http://aws.amazon.com/).

`spark-ec2` is designed to manage multiple named clusters. You can
launch a new cluster (telling the script its size and giving it a name),
shutdown an existing cluster, or log into a cluster. Each cluster
launches a set of instances, which are tagged with the cluster name,
and placed into EC2 security groups. If you don't specify a security
group, the `spark-ec2` script will create security groups based on the
cluster name you request. For example, a cluster named
shutdown an existing cluster, or log into a cluster. Each cluster is
identified by placing its machines into EC2 security groups whose names
are derived from the name of the cluster. For example, a cluster named
`test` will contain a master node in a security group called
`test-master`, and a number of slave nodes in a security group called
`test-slaves`. You can also specify a security group prefix to be used
in place of the cluster name. Machines in a cluster can be identified
by looking for the "Name" tag of the instance in the Amazon EC2 Console.
`test-slaves`. The `spark-ec2` script will create these security groups
for you based on the cluster name you request. You can also use them to
identify machines belonging to each cluster in the Amazon EC2 Console.


# Before You Start
Expand Down
73 changes: 21 additions & 52 deletions ec2/spark_ec2.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -47,7 +47,6 @@
AMI_PREFIX = "https://raw.github.com/mesos/spark-ec2/{b}/ami-list".format(b=MESOS_SPARK_EC2_BRANCH)



class UsageError(Exception):
pass

Expand Down Expand Up @@ -139,7 +138,7 @@ def parse_args():
help="The SSH user you want to connect as (default: %default)")
parser.add_option(
"--delete-groups", action="store_true", default=False,
help="When destroying a cluster, delete the security groups that were created.")
help="When destroying a cluster, delete the security groups that were created")
parser.add_option(
"--use-existing-master", action="store_true", default=False,
help="Launch fresh slaves, but use an existing stopped master if possible")
Expand All @@ -153,9 +152,6 @@ def parse_args():
parser.add_option(
"--user-data", type="string", default="",
help="Path to a user-data file (most AMI's interpret this as an initialization script)")
parser.add_option(
"--security-group-prefix", type="string", default=None,
help="Use this prefix for the security group rather than the cluster name.")
parser.add_option(
"--authorized-address", type="string", default="0.0.0.0/0",
help="Address to authorize on created security groups (default: %default)")
Expand Down Expand Up @@ -306,12 +302,8 @@ def launch_cluster(conn, opts, cluster_name):
user_data_content = user_data_file.read()

print "Setting up security groups..."
if opts.security_group_prefix is None:
master_group = get_or_make_group(conn, cluster_name + "-master")
slave_group = get_or_make_group(conn, cluster_name + "-slaves")
else:
master_group = get_or_make_group(conn, opts.security_group_prefix + "-master")
slave_group = get_or_make_group(conn, opts.security_group_prefix + "-slaves")
master_group = get_or_make_group(conn, cluster_name + "-master")
slave_group = get_or_make_group(conn, cluster_name + "-slaves")
authorized_address = opts.authorized_address
if master_group.rules == []: # Group was just now created
master_group.authorize(src_group=master_group)
Expand All @@ -336,11 +328,12 @@ def launch_cluster(conn, opts, cluster_name):
slave_group.authorize('tcp', 60060, 60060, authorized_address)
slave_group.authorize('tcp', 60075, 60075, authorized_address)

# Check if instances are already running with the cluster name
# Check if instances are already running in our groups
existing_masters, existing_slaves = get_existing_cluster(conn, opts, cluster_name,
die_on_error=False)
if existing_slaves or (existing_masters and not opts.use_existing_master):
print >> stderr, ("ERROR: There are already instances for name: %s " % cluster_name)
print >> stderr, ("ERROR: There are already instances running in " +
"group %s or %s" % (master_group.name, slave_group.name))
sys.exit(1)

# Figure out Spark AMI
Expand Down Expand Up @@ -414,13 +407,9 @@ def launch_cluster(conn, opts, cluster_name):
for r in reqs:
id_to_req[r.id] = r
active_instance_ids = []
outstanding_request_ids = []
for i in my_req_ids:
if i in id_to_req:
if id_to_req[i].state == "active":
active_instance_ids.append(id_to_req[i].instance_id)
else:
outstanding_request_ids.append(i)
if i in id_to_req and id_to_req[i].state == "active":
active_instance_ids.append(id_to_req[i].instance_id)
if len(active_instance_ids) == opts.slaves:
print "All %d slaves granted" % opts.slaves
reservations = conn.get_all_instances(active_instance_ids)
Expand All @@ -429,8 +418,8 @@ def launch_cluster(conn, opts, cluster_name):
slave_nodes += r.instances
break
else:
print "%d of %d slaves granted, waiting longer for request ids including %s" % (
len(active_instance_ids), opts.slaves, outstanding_request_ids[0:10])
print "%d of %d slaves granted, waiting longer" % (
len(active_instance_ids), opts.slaves)
except:
print "Canceling spot instance requests"
conn.cancel_spot_instance_requests(my_req_ids)
Expand Down Expand Up @@ -488,29 +477,14 @@ def launch_cluster(conn, opts, cluster_name):
print "Launched master in %s, regid = %s" % (zone, master_res.id)

# Give the instances descriptive names
# TODO: Add retry logic for tagging with name since it's used to identify a cluster.
for master in master_nodes:
name = '{cn}-master-{iid}'.format(cn=cluster_name, iid=master.id)
for i in range(0, 5):
try:
master.add_tag(key='Name', value=name)
except:
print "Failed attempt %i of 5 to tag %s" % ((i + 1), name)
if (i == 5):
raise "Error - failed max attempts to add name tag"
time.sleep(5)


master.add_tag(
key='Name',
value='{cn}-master-{iid}'.format(cn=cluster_name, iid=master.id))
for slave in slave_nodes:
name = '{cn}-slave-{iid}'.format(cn=cluster_name, iid=slave.id)
for i in range(0, 5):
try:
slave.add_tag(key='Name', value=name)
except:
print "Failed attempt %i of 5 to tag %s" % ((i + 1), name)
if (i == 5):
raise "Error - failed max attempts to add name tag"
time.sleep(5)
slave.add_tag(
key='Name',
value='{cn}-slave-{iid}'.format(cn=cluster_name, iid=slave.id))

# Return all the instances
return (master_nodes, slave_nodes)
Expand All @@ -528,19 +502,18 @@ def get_existing_cluster(conn, opts, cluster_name, die_on_error=True):
for res in reservations:
active = [i for i in res.instances if is_active(i)]
for inst in active:
name = inst.tags.get(u'Name', "")
if name.startswith(cluster_name + "-master"):
group_names = [g.name for g in inst.groups]
if group_names == [cluster_name + "-master"]:
master_nodes.append(inst)
elif name.startswith(cluster_name + "-slave"):
elif group_names == [cluster_name + "-slaves"]:
slave_nodes.append(inst)
if any((master_nodes, slave_nodes)):
print "Found %d master(s), %d slaves" % (len(master_nodes), len(slave_nodes))
if master_nodes != [] or not die_on_error:
return (master_nodes, slave_nodes)
else:
if master_nodes == [] and slave_nodes != []:
print >> sys.stderr, "ERROR: Could not find master in with name " + \
cluster_name + "-master"
print >> sys.stderr, "ERROR: Could not find master in group " + cluster_name + "-master"
else:
print >> sys.stderr, "ERROR: Could not find any existing cluster"
sys.exit(1)
Expand Down Expand Up @@ -976,11 +949,7 @@ def real_main():
# Delete security groups as well
if opts.delete_groups:
print "Deleting security groups (this will take some time)..."
if opts.security_group_prefix is None:
group_names = [cluster_name + "-master", cluster_name + "-slaves"]
else:
group_names = [opts.security_group_prefix + "-master",
opts.security_group_prefix + "-slaves"]
group_names = [cluster_name + "-master", cluster_name + "-slaves"]
wait_for_cluster_state(
cluster_instances=(master_nodes + slave_nodes),
cluster_state='terminated',
Expand Down

0 comments on commit f0b708b

Please sign in to comment.