-
Bug
-
Resolution: Fixed
-
Major
-
None
-
EC2 Plugin v 1.29
Jenkins 1.624
-
Powered by SuggestiMate
In every 10 or so instance launches, I see cases where the EC2 plugin opens a SSH connection to the new instance before the time where I believe the corresponding private key has been put in place by the AWS infrastructure during the launch of the instance. As a result, we see errors in the node launch. The output for the launch is below.
just before slave CentOS (i-f295e51b) gets launched ...
executing pre-launch scripts ...
Node CentOS (i-f295e51b)(i-f295e51b) is still pending/launching, waiting 5s
Node CentOS (i-f295e51b)(i-f295e51b) is still pending/launching, waiting 5s
Node CentOS (i-f295e51b)(i-f295e51b) is still pending/launching, waiting 5s
Node CentOS (i-f295e51b)(i-f295e51b) is still pending/launching, waiting 5s
Node CentOS (i-f295e51b)(i-f295e51b) is still pending/launching, waiting 5s
Node CentOS (i-f295e51b)(i-f295e51b) is still pending/launching, waiting 5s
Node CentOS (i-f295e51b)(i-f295e51b) is ready
Connecting to 10.240.1.146 on port 22, with timeout 10000.
Failed to connect via ssh: The kexTimeout (10000 ms) expired.
Waiting for SSH to come up. Sleeping 5.
Connecting to 10.240.1.146 on port 22, with timeout 10000.
Failed to connect via ssh: The kexTimeout (10000 ms) expired.
Waiting for SSH to come up. Sleeping 5.
Connecting to 10.240.1.146 on port 22, with timeout 10000.
Failed to connect via ssh: The kexTimeout (10000 ms) expired.
Waiting for SSH to come up. Sleeping 5.
Connecting to 10.240.1.146 on port 22, with timeout 10000.
Connected via SSH.
bootstrap()
Getting keypair...
Using key: jenkins
dc:xx:xx:xx
----BEGIN RSA PRIVATE KEY----
MIIEow<private key info>
Authenticating as centos
Authentication failed. Trying again...
Authenticating as centos
Authentication failed. Trying again...
Authenticating as centos
Authentication failed. Trying again...
Authenticating as centos
Authentication failed. Trying again...
Authenticating as centos
Authentication failed. Trying again...
Authenticating as centos
ERROR: Publickey authentication failed.
java.io.IOException: Publickey authentication failed.
at com.trilead.ssh2.auth.AuthenticationManager.authenticatePublicKey(AuthenticationManager.java:315)
at com.trilead.ssh2.Connection.authenticateWithPublicKey(Connection.java:467)
at hudson.plugins.ec2.ssh.EC2UnixLauncher.bootstrap(EC2UnixLauncher.java:260)
at hudson.plugins.ec2.ssh.EC2UnixLauncher.launch(EC2UnixLauncher.java:91)
at hudson.plugins.ec2.EC2ComputerLauncher.launch(EC2ComputerLauncher.java:107)
at hudson.slaves.SlaveComputer$1.call(SlaveComputer.java:238)
at jenkins.util.ContextResettingExecutorService$2.call(ContextResettingExecutorService.java:46)
at java.util.concurrent.FutureTask.run(FutureTask.java:266)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
Caused by: java.io.IOException: The connection is closed.
at com.trilead.ssh2.auth.AuthenticationManager.deQueue(AuthenticationManager.java:63)
at com.trilead.ssh2.auth.AuthenticationManager.getNextMessage(AuthenticationManager.java:86)
at com.trilead.ssh2.auth.AuthenticationManager.authenticatePublicKey(AuthenticationManager.java:290)
... 10 more
Caused by: java.io.IOException: Peer sent DISCONNECT message (reason code 2): Too many authentication failures for centos
at com.trilead.ssh2.transport.TransportManager.receiveLoop(TransportManager.java:766)
at com.trilead.ssh2.transport.TransportManager$1.run(TransportManager.java:489)
... 1 more
- is duplicated by
-
JENKINS-21891 slave fails to connect if dns takes too long to provide name
-
- Closed
-
-
JENKINS-24691 importing own private/pub to AWS ec2 doesn't work
-
- Closed
-
[JENKINS-30284] EC2 plugin too aggressive in timing in contacting new AWS instance over SSH
Same issue here, running CentOS 7 slaves. Output looks the same as above please make these timing configurable! Thanks
Node Jenkins Docker Slave (i-386b0fc1)(i-386b0fc1) is still pending/launching, waiting 5s Node Jenkins Docker Slave (i-386b0fc1)(i-386b0fc1) is still pending/launching, waiting 5s Node Jenkins Docker Slave (i-386b0fc1)(i-386b0fc1) is still pending/launching, waiting 5s Node Jenkins Docker Slave (i-386b0fc1)(i-386b0fc1) is still pending/launching, waiting 5s Node Jenkins Docker Slave (i-386b0fc1)(i-386b0fc1) is still pending/launching, waiting 5s Node Jenkins Docker Slave (i-386b0fc1)(i-386b0fc1) is still pending/launching, waiting 5s Node Jenkins Docker Slave (i-386b0fc1)(i-386b0fc1) is ready Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: There was a problem while connecting to ec2-52-91-128-246.compute-1.amazonaws.com:22 Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: There was a problem while connecting to ec2-52-91-128-246.compute-1.amazonaws.com:22 Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: There was a problem while connecting to ec2-52-91-128-246.compute-1.amazonaws.com:22 Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: There was a problem while connecting to ec2-52-91-128-246.compute-1.amazonaws.com:22 Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: There was a problem while connecting to ec2-52-91-128-246.compute-1.amazonaws.com:22 Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Failed to connect via ssh: The kexTimeout (10000 ms) expired. Waiting for SSH to come up. Sleeping 5. Connecting to ec2-52-91-128-246.compute-1.amazonaws.com on port 22, with timeout 10000. Connected via SSH. bootstrap() Getting keypair... Using key: np-dev-key 92:xx:yy:zz -----BEGIN RSA PRIVATE KEY----- MIIEoXXXX<snip> Authenticating as centos Authentication failed. Trying again... Authenticating as centos Authentication failed. Trying again... Authenticating as centos Authentication failed. Trying again... Authenticating as centos Authentication failed. Trying again... Authenticating as centos Authentication failed. Trying again... Authenticating as centos ERROR: Publickey authentication failed. java.io.IOException: Publickey authentication failed. at com.trilead.ssh2.auth.AuthenticationManager.authenticatePublicKey(AuthenticationManager.java:315) at com.trilead.ssh2.Connection.authenticateWithPublicKey(Connection.java:467) at hudson.plugins.ec2.ssh.EC2UnixLauncher.bootstrap(EC2UnixLauncher.java:260) at hudson.plugins.ec2.ssh.EC2UnixLauncher.launch(EC2UnixLauncher.java:91) at hudson.plugins.ec2.EC2ComputerLauncher.launch(EC2ComputerLauncher.java:107) at hudson.slaves.SlaveComputer$1.call(SlaveComputer.java:253) at jenkins.util.ContextResettingExecutorService$2.call(ContextResettingExecutorService.java:46) at java.util.concurrent.FutureTask.run(FutureTask.java:266) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) Caused by: java.io.IOException: The connection is closed. at com.trilead.ssh2.auth.AuthenticationManager.deQueue(AuthenticationManager.java:63) at com.trilead.ssh2.auth.AuthenticationManager.getNextMessage(AuthenticationManager.java:86) at com.trilead.ssh2.auth.AuthenticationManager.authenticatePublicKey(AuthenticationManager.java:290) ... 10 more Caused by: java.io.IOException: Peer sent DISCONNECT message (reason code 2): Too many authentication failures for centos at com.trilead.ssh2.transport.TransportManager.receiveLoop(TransportManager.java:766) at com.trilead.ssh2.transport.TransportManager$1.run(TransportManager.java:489) ... 1 more
This seems to be getting worse - maybe extra load on AWS. Even having the ability to tweak these timeouts would be very helpful. I'm manually killing off 10 - 15 slaves a day now.
Currently the timeout for this is set at 10000ms (10 seconds). Should the default be more? We can certainly put a mechanism in to make this configurable.
Looking more closely at this, I'm not 100% sure it's related to a "timeout", but maybe a chicken and egg situation where AWS hasn't dropped the SSH key yet. Would it be possible to retry that SSH some configurable number of times if it fails?
Below is our code. And it looks like the problem is we are retrying too frequently and have 5 failures and centos does not like this. But I'm not sure what we are waiting for to change on the slave so that the authentication will eventually succeed. I mean 10 seconds between each try seems like a pretty long time.
Any ideas of what it's waiting for?
private int bootstrap(Connection bootstrapConn, EC2Computer computer, PrintStream logger) throws IOException, InterruptedException, AmazonClientException { logger.println("bootstrap()"); boolean closeBootstrap = true; try { int tries = 20; boolean isAuthenticated = false; logger.println("Getting keypair..."); KeyPair key = computer.getCloud().getKeyPair(); logger.println("Using key: " + key.getKeyName() + "\n" + key.getKeyFingerprint() + "\n" + key.getKeyMaterial().substring(0, 160)); while (tries-- > 0) { logger.println("Authenticating as " + computer.getRemoteAdmin()); isAuthenticated = bootstrapConn.authenticateWithPublicKey(computer.getRemoteAdmin(), key.getKeyMaterial().toCharArray(), ""); if (isAuthenticated) { break; } logger.println("Authentication failed. Trying again..."); Thread.sleep(10000); } if (!isAuthenticated) { logger.println("Authentication failed"); return FAILED; } closeBootstrap = false; return SAMEUSER; } finally { if (closeBootstrap) bootstrapConn.close(); } }
I'm not totally sure, but it looks like I see OpenSSH start
Jan 7 19:39:14 ip-10-1-1-219 systemd: Started OpenSSH server daemon.
and then get restarted later by cloud-init when the authorized_keys file is put in place.
Jan 7 19:39:31 ip-10-1-1-219 systemd: Started Initial cloud-init job (metadata service crawler). Jan 7 19:39:31 ip-10-1-1-219 systemd: Starting Cloud-config availability. Jan 7 19:39:31 ip-10-1-1-219 systemd: Reached target Cloud-config availability. Jan 7 19:39:31 ip-10-1-1-219 systemd: Starting Apply the settings specified in cloud-config... Jan 7 19:39:32 ip-10-1-1-219 systemd: Stopping OpenSSH server daemon... Jan 7 19:39:32 ip-10-1-1-219 systemd: Started OpenSSH Server Key Generation. Jan 7 19:39:32 ip-10-1-1-219 systemd: Starting OpenSSH server daemon... Jan 7 19:39:32 ip-10-1-1-219 systemd: Started OpenSSH server daemon. Jan 7 19:39:32 ip-10-1-1-219 systemd: Started Apply the settings specified in cloud-config. Jan 7 19:39:32 ip-10-1-1-219 systemd: Starting Execute cloud user/final scripts... Jan 7 19:39:32 ip-10-1-1-219 ec2: Jan 7 19:39:32 ip-10-1-1-219 ec2: ############################################################# Jan 7 19:39:32 ip-10-1-1-219 ec2: -----BEGIN SSH HOST KEY FINGERPRINTS----- Jan 7 19:39:32 ip-10-1-1-219 ec2: xxxxxxxx (ECDSA) Jan 7 19:39:32 ip-10-1-1-219 ec2: xxxxxxxx (ED25519) Jan 7 19:39:32 ip-10-1-1-219 ec2: xxxxxxxx (RSA) Jan 7 19:39:32 ip-10-1-1-219 ec2: -----END SSH HOST KEY FINGERPRINTS----- Jan 7 19:39:32 ip-10-1-1-219 ec2: ############################################################# Jan 7 19:39:33 ip-10-1-1-219 systemd: Started Execute cloud user/final scripts.
Maybe you are starting your authentication attempts during that first phase of SSH being up which eventually times out or centos says 'go away'.
EDIT: the above logs are from a slave that succeeded by the way, so those times could be much different in the failure case.
Thanks, that's helpful. Any way of getting the logs on both sides from the failure case (of the same failure)?
Sure, I'll collect logs on the next failure. Unfortunately, it looks like the slave.log's don't have timestamps.
So it's a little tricky to wade through the log output, but this (I think) illustrates my thought (from a failed slave):
Initial OpenSSH startup on slave:
Jan 7 20:33:08 ip-10-1-1-219 systemd: Starting OpenSSH server daemon... Jan 7 20:33:09 ip-10-1-1-219 systemd: Started OpenSSH server daemon.
Slave failure in jenkins.log:
Jan 07, 2016 8:34:31 PM hudson.slaves.NodeProvisioner$2 run WARNING: Provisioned slave Jenkins Docker Slave (ami-8ad0a2e0) failed to launch java.util.concurrent.ExecutionException: java.io.IOException: Slave failed to connect, even though the launcher didn't report it. See the log output for details. at java.util.concurrent.FutureTask.report(FutureTask.java:122) at java.util.concurrent.FutureTask.get(FutureTask.java:192) at hudson.plugins.ec2.EC2Cloud$1.call(EC2Cloud.java:409) at hudson.plugins.ec2.EC2Cloud$1.call(EC2Cloud.java:394) at jenkins.util.ContextResettingExecutorService$2.call(ContextResettingExecutorService.java:46) at java.util.concurrent.FutureTask.run(FutureTask.java:266) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) Caused by: java.io.IOException: Slave failed to connect, even though the launcher didn't report it. See the log output for details. at hudson.slaves.SlaveComputer$1.call(SlaveComputer.java:277) ... 5 more Jan 07, 2016 8:34:41 PM hudson.plugins.ec2.EC2Cloud provision
Cloud-init on slave:
Jan 7 20:34:47 ip-10-1-1-219 systemd: Started Initial cloud-init job (metadata service crawler). Jan 7 20:34:47 ip-10-1-1-219 systemd: Starting Cloud-config availability. Jan 7 20:34:47 ip-10-1-1-219 systemd: Reached target Cloud-config availability. Jan 7 20:34:47 ip-10-1-1-219 systemd: Starting Apply the settings specified in cloud-config... Jan 7 20:34:51 ip-10-1-1-219 systemd: Stopping OpenSSH server daemon... Jan 7 20:34:51 ip-10-1-1-219 systemd: Started OpenSSH Server Key Generation. Jan 7 20:34:51 ip-10-1-1-219 systemd: Starting OpenSSH server daemon... Jan 7 20:34:51 ip-10-1-1-219 systemd: Started OpenSSH server daemon. Jan 7 20:34:51 ip-10-1-1-219 systemd: Started Apply the settings specified in cloud-config. Jan 7 20:34:51 ip-10-1-1-219 systemd: Starting Execute cloud user/final scripts... Jan 7 20:34:52 ip-10-1-1-219 ec2: Jan 7 20:34:52 ip-10-1-1-219 ec2: ############################################################# Jan 7 20:34:52 ip-10-1-1-219 ec2: -----BEGIN SSH HOST KEY FINGERPRINTS----- Jan 7 20:34:53 ip-10-1-1-219 ec2: 256 xxxxxxxx (ECDSA) Jan 7 20:34:53 ip-10-1-1-219 ec2: 256 xxxxxxxx (ED25519) Jan 7 20:34:53 ip-10-1-1-219 ec2: 2048 xxxxxxxx (RSA) Jan 7 20:34:53 ip-10-1-1-219 ec2: -----END SSH HOST KEY FINGERPRINTS----- Jan 7 20:34:53 ip-10-1-1-219 ec2: ############################################################# Jan 7 20:34:53 ip-10-1-1-219 systemd: Started Execute cloud user/final scripts. Jan 7 20:34:53 ip-10-1-1-219 systemd: Starting Multi-User System. Jan 7 20:34:53 ip-10-1-1-219 systemd: Reached target Multi-User System. Jan 7 20:34:53 ip-10-1-1-219 systemd: Starting Update UTMP about System Runlevel Changes... Jan 7 20:34:53 ip-10-1-1-219 systemd: Started Stop Read-Ahead Data Collection 10s After Completed Startup. Jan 7 20:34:53 ip-10-1-1-219 systemd: Started Update UTMP about System Runlevel Changes. Jan 7 20:34:53 ip-10-1-1-219 systemd: Startup finished in 1.528s (kernel) + 35.815s (initrd) + 3min 18.160s (userspace) = 3min 55.504s.
Hi -
Sorry to be a pest, but is there any forward motion here? I'm going crazy killing dead slaves.... Thanks!
-Camden
Hey Camden,
Would you be OK if I gave you a couple of system properties that you can set to control the time to wait and the number of times to loop? And also I fixed the logging so that it has timestamps on it? I can put this in today and you can take a snapshot build (let me know if you need help to get the hpi file) and try it out to see if it fixes your problem before I release it. If it's truly the timeout value that sometimes needs to be different depending on the OS or such, then I can move it form a system property to something in the UI, but let's confirm that's the case first.
Let me know.
Francis
Yeah that works for me. Let me know how to find the hpi. Thanks!
-Camden
Code changed in jenkins
User: Francis Upton IV
Path:
src/main/java/hudson/plugins/ec2/EC2AbstractSlave.java
src/main/java/hudson/plugins/ec2/EC2Cloud.java
src/main/java/hudson/plugins/ec2/EC2Computer.java
src/main/java/hudson/plugins/ec2/EC2ComputerLauncher.java
src/main/java/hudson/plugins/ec2/ssh/EC2UnixLauncher.java
http://jenkins-ci.org/commit/ec2-plugin/86f1defc1edb6c85f6e0a1ab99c969e875148f72
Log:
JENKINS-32439 Incorrect slave template (AMI) found when launching slave
JENKINS-30284 EC2 plugin too aggressive in timing in contacting new AWS instance over SSH (parameterized timeouts, added logging)
You can get the hpi here.
The properties are (with their default values [these were the existing values])
private static final String BOOTSTRAP_AUTH_SLEEP_MS = "jenkins.ec2.bootstrapAuthSleepMs";
private static final String BOOTSTRAP_AUTH_TRIES= "jenkins.ec2.bootstrapAuthTries";
private static int bootstrapAuthSleepMs = 10000;
private static int bootstrapAuthTries = 20;
Thanks Francis - we are now running the snapshot. I'll let you know how it goes!
Hi Francis - Just a quick update. Bumping the sleep to 30000 and the tries to 30 has completely solved my problem so far. Cheers!
Hi again -
It's been a few days now and this has completely resolved the issues we were seeing. I would :+1: some UI around these options and allowing them to be set.
Cheers,
-Camden
Code changed in jenkins
User: Francis Upton IV
Path:
src/main/java/hudson/plugins/ec2/ssh/EC2UnixLauncher.java
http://jenkins-ci.org/commit/ec2-plugin/06c7369977c01520b1d89a3a868d779cdd2600c9
Log:
JENKINS-30284 EC2 plugin too aggressive in timing in contacting new AWS instance over SSH (changed default timeouts)
I think I'm going to change the defaults to 30000/30 and leave the properties as they are. There is already a lot of obscure UI around various things and I think since this can be tweaked by the properties if necessary (and maybe it won't even be necessary with better defaults) I'm reluctant to expose it in the UI. If people really want then, I can do it, but let's wait and see on that. I will document the properties on the Wiki.
we're having an issue with the ssh client connecting before user data is done.
how do we edit bootstrapAuthSleepMs?
UPDATE: in case anyone else needs to know add -Djenkins.ec2.bootstrapAuthSleepMs=60000 to your jenkins launch command (java -jar jenkins.war -D...) to add a 60 second wait.
We are running into the same problem with a CentOS slave (not sure if the OS matters) and it seems to be approximately 1 in 10 or so. Our output looks the same as the one from the description above.
When that happens the affected slave just says "Ping response time is too long or timed out." and hangs around until we delete it.