Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

SPARK-2333 - spark_ec2 script should allow option for existing security group #1899

Closed
wants to merge 2 commits into from
+57 −28
Closed

SPARK-2333 - spark_ec2 script should allow option for existing security group #1899

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
b2989d5
SPARK-2333: spark_ec2 script should allow option for existing securit…
vidaha Aug 12, 2014
c80d5c3
wrap retries in a try catch block
vidaha Aug 19, 2014
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 8 additions & 6 deletions docs/ec2-scripts.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,14 +12,16 @@ 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 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
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
`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`. 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.
`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.


# Before You Start
Expand Down
64 changes: 42 additions & 22 deletions ec2/spark_ec2.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -124,7 +124,7 @@ def parse_args():
help="The SSH user you want to connect as (default: root)")
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 @@ -138,7 +138,9 @@ 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.")

(opts, args) = parser.parse_args()
if len(args) != 2:
Expand Down Expand Up @@ -285,8 +287,12 @@ def launch_cluster(conn, opts, cluster_name):
user_data_content = user_data_file.read()

print "Setting up security groups..."
master_group = get_or_make_group(conn, cluster_name + "-master")
slave_group = get_or_make_group(conn, cluster_name + "-slaves")
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")
if master_group.rules == []: # Group was just now created
master_group.authorize(src_group=master_group)
master_group.authorize(src_group=slave_group)
Expand All @@ -310,12 +316,11 @@ def launch_cluster(conn, opts, cluster_name):
slave_group.authorize('tcp', 60060, 60060, '0.0.0.0/0')
slave_group.authorize('tcp', 60075, 60075, '0.0.0.0/0')

# Check if instances are already running in our groups
# Check if instances are already running with the cluster name
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 running in " +
"group %s or %s" % (master_group.name, slave_group.name))
print >> stderr, ("ERROR: There are already instances for name: %s " % cluster_name)
sys.exit(1)

# Figure out Spark AMI
Expand Down Expand Up @@ -371,9 +376,13 @@ 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 and id_to_req[i].state == "active":
active_instance_ids.append(id_to_req[i].instance_id)
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 len(active_instance_ids) == opts.slaves:
print "All %d slaves granted" % opts.slaves
reservations = conn.get_all_instances(active_instance_ids)
Expand All @@ -382,8 +391,8 @@ def launch_cluster(conn, opts, cluster_name):
slave_nodes += r.instances
break
else:
print "%d of %d slaves granted, waiting longer" % (
len(active_instance_ids), opts.slaves)
print "%d of %d slaves granted, waiting longer for request ids including %s" % (
len(active_instance_ids), opts.slaves, outstanding_request_ids[0:10])
except:
print "Canceling spot instance requests"
conn.cancel_spot_instance_requests(my_req_ids)
Expand Down Expand Up @@ -440,14 +449,22 @@ 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.
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Minor nit: You can remove this TODO now.

for master in master_nodes:
master.add_tag(
key='Name',
value='{cn}-master-{iid}'.format(cn=cluster_name, iid=master.id))
name = '{cn}-master-{iid}'.format(cn=cluster_name, iid=master.id)
for i in range(0, 5):
master.add_tag(key='Name', value=name)
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What happens if an add_tag call fails? My bet is that it throws an exception rather than silently failing, in which case this re-try logic won't run. Rather than using this "set-and-test" logic, maybe we can just wrap the call in a try-except block?

@shivaram Did the eventual-consistency issue that you saw result in exceptions from add_tag?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes - I am pretty sure it throws an exception. I don't remember what the type is -- All I see in my notes is that the exception says 'Instance not found'

if master.tags.get(u'Name', "") == name:
break
print "Failed attempt %i of 5 to tag %s" % ((i + 1), name)

for slave in slave_nodes:
slave.add_tag(
key='Name',
value='{cn}-slave-{iid}'.format(cn=cluster_name, iid=slave.id))
name = '{cn}-slave-{iid}'.format(cn=cluster_name, iid=slave.id)
for i in range(0, 5):
slave.add_tag(key='Name', value=name)
if slave.tags.get(u'Name', "") == name:
break
print "Failed attempt %i of 5 to tag %s" % ((i + 1), name)

# Return all the instances
return (master_nodes, slave_nodes)
Expand All @@ -463,18 +480,18 @@ 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:
group_names = [g.name for g in inst.groups]
if group_names == [cluster_name + "-master"]:
name = inst.tags.get(u'Name', "")
if name.startswith(cluster_name + "-master"):
master_nodes.append(inst)
elif group_names == [cluster_name + "-slaves"]:
elif name.startswith(cluster_name + "-slave"):
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 group " + cluster_name + "-master"
print >> sys.stderr, "ERROR: Could not find master in with name " + cluster_name + "-master"
else:
print >> sys.stderr, "ERROR: Could not find any existing cluster"
sys.exit(1)
Expand Down Expand Up @@ -816,7 +833,10 @@ def real_main():
# Delete security groups as well
if opts.delete_groups:
print "Deleting security groups (this will take some time)..."
group_names = [cluster_name + "-master", cluster_name + "-slaves"]
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"]

attempt = 1
while attempt <= 3:
Expand Down