Skip to content
This repository was archived by the owner on Nov 26, 2024. It is now read-only.

recover_missing_vm.md

Latest commit

 

History

History
179 lines (141 loc) · 8.3 KB

recover_missing_vm.md

File metadata and controls

179 lines (141 loc) · 8.3 KB

Recover the missing vm from 'bosh deploy' failure

Table of Contents

Purpose

When bosh deploy fails, sometimes the failing vm is missing from bosh vms (deleted from bosh db) but still alive in Softlayer. If you do nothing, in the next bosh deploy, bosh will treat it as a missing vm and create a new one with different IP. Sometimes we need to keep the vm IP unchanged, so the missing vm needs to be recovered in the bosh db.

User Impact

If you don't perform this run book, the next bosh deploy will create a new vm to backfill the missing vm with different IP. This will break the cases when the vm IP needs to be kept unchanged.

Instructions to Fix

  • Scenario 1: user_data.json exists

    Make sure the failing vm can still been seen from Softlayer Portal and /var/vcap/bosh/user_data.json exists on the failing vm.

    Take nats/0 as an example of the failing vm.

    1) Update the vm info in "instances" table of bosh db

    update instances set vm_cid='12345678' where job='nats' and index=0;
update instances set agent_id='1234abcd-12ab-34cd-56ef-123456abcdef' where job='nats' and index=0;

    The vm_cid can be got from Softlayer Portal address such as 12345678 in https://control.softlayer.com/devices/details/12345678/virtualGuest

    The agent_id can be got from /var/vcap/bosh/user_data.json on the failing vm.

    2) Create spec.json for the failing vm

    Login to the failing vm and create /var/vcap/bosh/spec.json like this:

    {
  "job": {
    "name": "REPLACE_job_name"  
  },
  "deployment": "REPLACE_deployment_name",
  "networks": {
    "default": {
      "cloud_properties": {
        "security_groups": [
          "default",
          "cf"
        ]
      },
      "default": [
        "dns",
        "gateway"
      ],
      "dns": [
        "REPLACE_dns_ip",
        "10.0.80.11",
        "10.0.80.12"
      ],
      "gateway": "REPLACE_gateway_ip",
      "ip": "REPLACE_vm_ip",
      "netmask": "REPLACE_netmask_ip",
      "type": "dynamic"
    }
  },
  "index": REPLACE_job_index,
  "id": "",
  "persistent_disk": 0
}
    • name: Replace with the job name. It's nats in the example.
    • dns: Replace with the dns ip. In case the PowerDNS on director is used, it's the director ip.
    • gateway: Replace with the gateway ip. It can be found from route -n or netstat -rn. See example below.
    • ip: Replace with the vm ip. It can be found from ifconfig. See example below.
    • netmask: Replace with the netmask ip. It can be found from ifconfig. See example below.
    • deployment: Replace with the name of deployment which the vm belongs to.
    • index: Replace with the job index. It's 0 in the example.

    netstat -rn or route -n output example:

    Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
0.0.0.0         <gateway_ip>    0.0.0.0         UG    0      0        0 eth1
9.0.0.0         10.121.120.193  255.0.0.0       UG    0      0        0 eth0
10.0.0.0        10.121.120.193  255.0.0.0       UG    0      0        0 eth0
10.121.120.192  0.0.0.0         255.255.255.192 U     0      0        0 eth0
161.26.0.0      10.121.120.193  255.255.0.0     UG    0      0        0 eth0
169.53.1.128    0.0.0.0         255.255.255.224 U     0      0        0 eth1

    ifconfig output example:

    eth0      Link encap:Ethernet  HWaddr 06:99:32:27:34:f7
          inet addr:<vm_private_ip>  Bcast:10.121.120.255  Mask:<netmask_ip_private>
          inet6 addr: fe80::499:32ff:fe27:34f7/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:596214 errors:0 dropped:0 overruns:0 frame:0
          TX packets:5065 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:63759564 (63.7 MB)  TX bytes:661260 (661.2 KB)
 
eth1      Link encap:Ethernet  HWaddr 06:da:3f:89:2a:b2
          inet addr:<vm_public_ip>  Bcast:169.53.1.159  Mask:<netmask_ip_public>
          inet6 addr: fe80::4da:3fff:fe89:2ab2/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:65584 errors:0 dropped:0 overruns:0 frame:0
          TX packets:21240 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:5413249 (5.4 MB)  TX bytes:3236947 (3.2 MB)
 
lo        Link encap:Local Loopback
          inet addr:127.0.0.1  Mask:255.0.0.0
          inet6 addr: ::1/128 Scope:Host
          UP LOOPBACK RUNNING  MTU:65536  Metric:1
          RX packets:6291 errors:0 dropped:0 overruns:0 frame:0
          TX packets:6291 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:892748 (892.7 KB)  TX bytes:892748 (892.7 KB)

    3) Restart bosh agent on the failing vm

    sv restart agent

    4) Make sure no problem with "bosh vms" and "bosh cck"

    • bosh vms: the missing vm is back
    • bosh cck: no problems

    5) Re-run "bosh deploy"

  • Scenario 2: user_data.json does not exist

    Sometimes when bosh deploy fails, user_data.json has not yet been generated in cases like create_vm failure.

    Make sure the failing vm can still been seen from Softlayer Portal

    1) Do the steps 1) & 2) in Scenario 1

    Except the agent_id needs need to be got from the bosh debug log for the failure.

    bosh debug log example:

    E, [2016-07-22 06:17:15 #13984] [canary_update(nats/0 (14109da6-cc51-49e8-8e99-d096e4f59550))] ERROR -- DirectorJobRunner: error creating vm: Creating Virtual_Guest with agent ID '1234abcd-12ab-34cd-56ef-123456abcdef': Attaching ephemeral disk to VirtualGuest `12345678`: Waiting for VirtualGuest `12345678` has Service Setup transaction complete: Getting Last Complete Transaction for virtual guest with ID '12345678': Get https://[user]:[api_key]@api.softlayer.com/rest/v3/SoftLayer_Virtual_Guest/12345678/getLastTransaction.json?objectMask=transactionGroup: net/http: TLS handshake timeout

    2) Create /var/vcap/bosh/user_data.json

    user_data.json example:

    {"agent_id":"1234abcd-12ab-34cd-56ef-123456abcdef","vm":{"name":"vm-1234abcd-12ab-34cd-56ef-123456abcdef","id":"vm-1234abcd-12ab-34cd-56ef-123456abcdef"},"mbus":"nats://nats:nats@[director_ip]:4222","ntp":[],"blobstore":{"provider":"dav","options":{"endpoint":"http://[director_ip]:25250","password":"[agent_password]","user":"agent"}},"networks":{"default":{"type":"dynamic","ip":"[vm_ip]","netmask":"[netmask]","gateway":"[gateway_ip]","dns":["[dns_ip]","10.0.80.11","10.0.80.12"],"default":["dns","gateway"],"preconfigured":true,"cloud_properties":{"security_groups":["default","cf"]}}},"disks":{"ephemeral":"/dev/xvdc","persistent":{}},"env":{}}

    Update the items in bold:

    • agent_id can be found in the bosh debug log
    • Keep "persistent" as empty which will be fixed later)
    • Refer to Scenario 1 to get vm_ip, netmask and gateway_ip

    3) Restart bosh agent

    sv restart agent

    4) Run bosh cck and fix the persistent disk problem if exists

    Problem 1 of 1: Inconsistent mount information:
Record shows that disk '12661853' should be mounted on 22354779.   
However it is currently :    
    Not mounted in any VM.    
  1. Ignore    
  2. Reattach disk to instance    
  3. Reattach disk and reboot instance    
Please choose a resolution [1 - 3]: 2

    5) Make sure no problem with "bosh vms" and "bosh cck"

    • bosh vms: the missing vm is back
    • bosh cck: no problems

    6) Re-run bosh deploy