This page describes troubleshooting steps that you might find helpful if you run into problems using Compute Engine instances.
Troubleshooting general issues with instances
If your instance doesn't start up
Here are some tips to help troubleshoot your persistent boot disk if it doesn't boot.
Your boot disk is full. If your boot disk is completely full and your operating system does not support automatic resizing, you won't be able to SSH into your instance. You must create a new instance and recreate the boot disk. See recovering an inaccessible instance or a full boot disk.
Examine your virtual machine instance's serial port output.
An instance's BIOS, bootloader, and kernel will print their debug messages into the instance's serial port output, providing valuable information about any errors or issues that the instance experienced. If you enable serial port output logging to Stackdriver, you can access this information even when your instance is not running. See viewing serial port output.
Enable interactive access to the serial console.
You can enable interactive access to an instance's serial console so you can log in and debug boot issues from within the instance, without requiring your instance to be fully booted. For more information, read Interacting with the serial console.
Verify that your disk has a valid file system.
If your file system is corrupted or otherwise invalid, you won't be able to launch your instance. Validate your disk's file system:
Detach the disk in question from any instance it is attached to, if applicable:
gcloud compute instances delete old-instance --keep-disks boot
Start a new instance with the latest Google-provided image:
gcloud compute instances create debug-instance
Attach your disk as a non-boot disk but don't mount it. Replace
DISKwith the name of the disk that won't boot. Note that we also provide a device name so that the disk is easily identifiable on the instance:
gcloud compute instances attach-disk debug-instance --disk DISK --device-name debug-disk
Connect to the instance:
gcloud compute ssh debug-instance
Look up the root partition of the disk, which is identified with the
part1notation. In this case, the root partition of the disk is at
user@debug-instance:~$ ls -l /dev/disk/by-id total 0 lrwxrwxrwx 1 root root 9 Jan 22 17:09 google-debug-disk -> ../../sdb lrwxrwxrwx 1 root root 10 Jan 22 17:09 google-debug-disk-part1 -> ../../sdb1 lrwxrwxrwx 1 root root 9 Jan 22 17:02 google-persistent-disk-0 -> ../../sda lrwxrwxrwx 1 root root 10 Jan 22 17:02 google-persistent-disk-0-part1 -> ../../sda1 lrwxrwxrwx 1 root root 9 Jan 22 17:09 scsi-0Google_PersistentDisk_debug-disk -> ../../sdb lrwxrwxrwx 1 root root 10 Jan 22 17:09 scsi-0Google_PersistentDisk_debug-disk-part1 -> ../../sdb1 lrwxrwxrwx 1 root root 9 Jan 22 17:02 scsi-0Google_PersistentDisk_persistent-disk-0 -> ../../sda lrwxrwxrwx 1 root root 10 Jan 22 17:02 scsi-0Google_PersistentDisk_persistent-disk-0-part1 -> ../../sda1
Run a file system check on the root partition:
user@debug-instance:~$ sudo fsck /dev/sdb1 fsck from util-linux 2.20.1 e2fsck 1.42.5 (29-Jul-2012) /dev/sdb1: clean, 19829/655360 files, 208111/2621184 blocks
Mount your file system:
user@debug-instance:~$ sudo mkdir /mydisk
user@debug-instance:~$ sudo mount /dev/sdb1 /mydisk
Check that the disk has kernel files:
user@debug-instance~:$ ls /mydisk/boot/vmlinuz-* /mydisk/boot/vmlinuz-3.2.0-4-amd64
Verify that the disk has a valid master boot record (MBR).
Run the following command on the debug instance that has attached the persistent boot disk, such as
$ sudo parted /dev/sdb print
If your MBR is valid, it should list information about the filesystem:
Disk /dev/sdb: 10.7GB Sector size (logical/physical): 512B/4096B Partition Table: msdos Disk Flags: Number Start End Size Type File system Flags 1 2097kB 10.7GB 10.7GB primary ext4 boot
If you cannot create an instance
Here are some tips to help troubleshoot if your instance is not created.
Simultaneous resource mutation or creation operations might cause an error. For example, if you are modifying secondary ranges in a subnetwork and creating a VM at the same time, you might see the following error:
The resource 'projects/[PROJECT]/regions/[REGION]/subnetworks/default' is not ready. The solution is to retry the failed operation.
If you receive a resource error (such as
ZONE_RESOURCE_POOL_EXHAUSTED_WITH_DETAILS) when requesting new resources, it means that the zone cannot currently accommodate your request. This error is due to Compute Engine resource obtainability, and is not due to your Compute Engine quota.
Here are some tips to help mitigate:
- Because this situation is temporary and can change frequently based on fluctuating demand, try your request again later.
- If possible, try to create the resources in another zone in the region or in another region.
- If possible, change the shape of the VM you are requesting. It's easier to get smaller machine types than larger ones. A change to your request, such as reducing the number of GPUs or using a custom VM with less memory or vCPUs, might allow your request to proceed.
- Use Compute Engine reservations to reserve resources within a zone to ensure that the resources you need are available when you need them.
- If you are trying to create a preemptible instance, remember that preemptible VMs are spare capacity and so might not be obtainable at peak demand periods.
- If you receive a
does not exist in zoneerror when requesting new resources, it means that the zone does not offer the resource or machine type that you have requested. See Regions and zones to find out which features are available in each zone.
If network traffic to or from your instance is being dropped
Compute Engine only allows network traffic that is explicitly permitted by your project's Firewall rules to reach your instance. By default, all projects automatically come with a default network that allows certain kinds of connections. If you deny all traffic, by default, that also denies SSH connections and all internal traffic. For more information, see the Firewall rules page.
In addition, you may need to adjust TCP keep-alive settings to work around the default idle connection timeout of 10 minutes. For more information, see Communicating between your instances and the internet.
Troubleshooting firewall rules or routes on an instance
The Google Cloud Platform Console provides network details for each network interface of an instance. You can view all of the firewall rules or routes that apply to an interface, or you can view just the rules and routes that the interface uses. Either view can help you troubleshoot which firewall rules and routes apply to the instance and which ones are actually being used (where priority and processing order override other rules or routes).
For more information, see the troubleshooting information in the Virtual Private Cloud documentation:
Troubleshooting issues with SSH
Under certain conditions, it is possible that a Linux instance
no longer accepts SSH connections. There are many reasons this could happen,
from a full disk to an accidental misconfiguration of
sshd. This section
describes a number of tips and approaches to troubleshoot and resolve common
Check your firewall rules
Compute Engine provisions each project with a default set of firewall
rules that permit SSH traffic. If the default firewall rule that permits SSH
connections is somehow removed, you'll be unable to access your instance. Check
your list of firewalls with the
gcloud command-line tool command-line tool and ensure
default-allow-ssh rule is present.
gcloud compute firewall-rules list
If it is missing, add it back:
gcloud compute firewall-rules create default-allow-ssh --allow tcp:22
Debug the issue in the serial console
You can enable read/write access to an instance's serial console so you can log into the console and troubleshoot problems with the instance. This is particularly useful when you cannot log in with SSH or if the instance has no connection to the network. The serial console remains accessible in both these conditions.
To learn how to enable interactive access and connect to an instance's serial console, read Interacting with the serial console.
Test the network
You can use the netcat tool to connect to your instance on port 22, and
determine whether the network connection is working. If you connect and see
an ssh banner (for example,
SSH-2.0-OpenSSH_6.0p1 Debian-4), your network
connection is working, and
you can rule out firewall problems. First, use the
gcloud tool to get
natIP for your instance:
gcloud compute instances describe example-instance --format='get(networkInterfaces.accessConfigs.natIP)' 198.51.100.8
nc command to connect to your instance:
# Check for SSH banner user@local:~$ nc [EXTERNAL_IP] 22 SSH-2.0-OpenSSH_6.0p1 Debian-4
Try a new user
The issue that prevents you from logging in might be limited to your account
(for example, if the permissions on the
~/.ssh/authorized_keys file on the
instance were set incorrectly).
Try using the
gcloud tool to log in as a fresh user, specifying
another username with the SSH request. The
gcloud tool will update the
project's metadata to add the new user and allow SSH access.
user@local:~$ gcloud compute ssh [USER]@example-instance
[USER] is a new username to log in with.
Use your disk on a new instance
If the above set of steps doesn't work for you, and the instance you're
interested in is booted from a persistent disk, you can detach the persistent
disk and attach this disk to use on the new instance.
DISK with your disk name:
gcloud compute instances delete old-instance --keep-disks=boot
gcloud compute instances create new-instance --disk name=DISK boot=yes auto-delete=no
gcloud compute ssh new-instance
Inspect an instance without shutting it down
You might have an instance you can't connect to that continues to correctly serve production traffic. In this case, you might want to inspect the disk without interrupting the instance's ability to serve users. First, take a snapshot of the instance's boot disk, then create a new disk from that snapshot, create a temporary instance, and finally attach and mount the new persistent disk to your temporary instance to troubleshoot the disk.
Create a new VPC network to host your cloned instance:
gcloud compute networks create debug-network
Add a firewall rule to allow SSH connections to the network:
gcloud compute firewall-rules create debug-network-allow-ssh --allow tcp:22
Create a snapshot of the disk in question, replacing
DISKwith the disk name:
gcloud compute disks snapshot DISK --snapshot-name debug-disk-snapshot
Create a new disk with the snapshot you just created:
gcloud compute disks create example-disk-debugging --source-snapshot debug-disk-snapshot
Create a new debugging instance without an external IP address:
gcloud compute instances create debugger --network debug-network --no-address
Attach the debugging disk to the instance:
gcloud compute instances attach-disk debugger --disk example-disk-debugging
Follow the instructions to connect to an instance without an external IP address.
After you are logged in to the debugger instance, troubleshoot the instance. For example, you can look at the instance logs:
$ sudo su -
$ mkdir /mnt/myinstance
$ mount /dev/disk/by-id/scsi-0Google_PersistentDisk_example-disk-debugging /mnt/myinstance
$ cd /mnt/myinstance/var/log
# Identify the issue preventing ssh from working $ ls
Use a startup script
If none of the earlier suggestions helped, you can create a startup script to collect information right after the instance starts. Follow the instructions for running a startup script.
Afterward, you need to reset your instance before the metadata takes affect
gcloud compute instances reset.
Alternatively, you can also recreate your instance with a diagnostic startup
gcloud compute instances deletewith the
gcloud compute instances delete INSTANCE --keep-disks boot
Add a new instance with the same disk and specify your startup script.
gcloud compute instances create example-instance --disk name=DISK,boot=yes --startup-script-url URL
As a starting point, you can use the compute-ssh-diagnostic script to collect diagnostics information for most common issues.