Skip to content
This repository has been archived by the owner on Sep 11, 2024. It is now read-only.

Latest commit

 

History

History
572 lines (490 loc) · 35.9 KB

Troubleshooting.md

File metadata and controls

572 lines (490 loc) · 35.9 KB
Raw

Troubleshooting Guide for OMS Agent for Linux

The following document provides a quick series of steps and procedures to diagnose and troubleshoot common issues. If none of these steps work for you the following channels for help are also available

Table of Contents

Important Log Locations and Log Collector Tool

File Path
OMS Agent for Linux Log File /var/opt/microsoft/omsagent/<workspace id>/log/omsagent.log
OMS Agent Configuration Log File /var/opt/microsoft/omsconfig/omsconfig.log

We recommend you to use our log collector tool to retrieve important logs for troubleshooting or before submitting a github issue. You can read more about the tool and how to run it here

Important Configuration Files

Category File Location
Syslog /etc/syslog-ng/syslog-ng.conf or /etc/rsyslog.conf or /etc/rsyslog.d/95-omsagent.conf
Performance, Nagios, Zabbix, OMS output and general agent /etc/opt/microsoft/omsagent/<workspace id>/conf/omsagent.conf
Additional configurations /etc/opt/microsoft/omsagent/<workspace id>/conf/omsagent.d/*.conf

Note: Editing configuration files for performance counters & syslog is overwritten if Portal Configuration is enabled. Disable configuration in the OMS Portal (all nodes) or for single nodes run the following:

sudo /opt/microsoft/omsconfig/Scripts/OMS_MetaConfigHelper.py --disable && sudo rm /etc/opt/omi/conf/omsconfig/configuration/Current.mof* /etc/opt/omi/conf/omsconfig/configuration/Pending.mof*

Installation Error Codes

Error Code Meaning
NOT_DEFINED Because the necessary dependencies are not installed, the auoms auditd plugin will not be installed
2 Invalid option provided to the shell bundle; Run sudo sh ./omsagent-*.universal*.sh --help for usage
3 No option provided to the shell bundle; Run sudo sh ./omsagent-*.universal*.sh --help for usage
4 Invalid package type OR invalid proxy settings; omsagent-rpm.sh packages can only be installed on RPM-based systems, and omsagent-deb.sh packages can only be installed on Debian-based systems; We recommend that you use the universal installer from the latest release. Also review your proxy settings.
5 The shell bundle must be executed as root OR there was 403 error returned during onboarding; Run your command using sudo
6 Invalid package architecture OR there was error 200 error returned during onboarding; omsagent-*x64.sh packages can only be installed on 64-bit systems, and omsagent-*x86.sh packages can only be installed on 32-bit systems; Download the correct package for your architecture from the latest release
17 Installation of OMS package failed; Look through the command output for the root failure
18 Installation of OMSConfig package failed; Look through the command output for the root failure; Any Github issues should be opened with the Powershell-DSC-for-Linux Github repo
19 Installation of OMI package failed; Look through the command output for the root failure; Any Github issues should be opened with the OMI Github repo
20 Installation of SCX package failed; Look through the command output for the root failure; Any Github issues should be opened with the SCXCore Github repo
21 Installation of Provider kits failed; Look through the command output for the root failure
22 Installation of bundled package failed; Look through the command output for the root failure
23 SCX or OMI package already installed; Use --upgrade instead of --install to install the shell bundle
30 Internal bundle error; File a GitHub Issue with details from the output
55 Unsupported openssl version OR Cannot connect to Microsoft OMS service OR dpkg is locked OR Missing curl program
61 Missing Python ctypes library; Install the Python ctypes library or package (python-ctypes)
62 Missing tar program; Install tar
63 Missing sed program; Install sed
64 Missing curl program; Install curl
65 Missing gpg program; Install gpg

Onboarding Error Codes

Error Code Meaning
2 Invalid option provided to the omsadmin script; Run sudo sh /opt/microsoft/omsagent/bin/omsadmin.sh -h for usage
3 Invalid configuration provided to the omsadmin script; Run sudo sh /opt/microsoft/omsagent/bin/omsadmin.sh -h for usage
4 Invalid proxy provided to the omsadmin script; Verify the proxy and see our documentation for using an HTTP proxy
5 403 HTTP error received from OMS service; See the full output of the omsadmin script for details
6 Non-200 HTTP error received from OMS service; See the full output of the omsadmin script for details
7 Unable to connect to OMS service; See the full output of the omsadmin script for details
8 Error onboarding to OMS workspace; See the full output of the omsadmin script for details
30 Internal script error; File a GitHub Issue with details from the output
31 Error generating agent ID; File a GitHub Issue with details from the output
32 Error generating certificates; See the full output of the omsadmin script for details
33 Error generating metaconfiguration for omsconfig; File a GitHub Issue with details from the output
34 Metaconfiguration generation script not present; Retry onboarding with sudo sh /opt/microsoft/omsagent/bin/omsadmin.sh -w <OMS Workspace ID> -s <OMS Workspace Key>

Enable Debug Logging

OMS output plugin debug

FluentD allows for plugin specific logging levels allowing you to specify different log levels for inputs and outputs. To specify a different log level for OMS output edit the general agent configuration at /etc/opt/microsoft/omsagent/<workspace id>/conf/omsagent.conf:

In the OMS output plugin, near the bottom of the configuration file, change the log_level property from info to debug

<match oms.** docker.**>
 type out_oms
 log_level debug
 num_threads 5
 buffer_chunk_limit 5m
 buffer_type file
 buffer_path /var/opt/microsoft/omsagent/<workspace id>/state/out_oms*.buffer
 buffer_queue_limit 10
 flush_interval 20s
 retry_limit 6
 retry_wait 30s
 max_retry_wait 30m
</match>

Debug logging allows you to see batched uploads to the OMS Service seperated by type, number of data items, and time taken to send:

Example debug enabled log:

Success sending oms.nagios x 1 in 0.14s
Success sending oms.omi x 4 in 0.52s
Success sending oms.syslog.authpriv.info x 1 in 0.91s

Verbose output

Instead of using the OMS output plugin you can also output Data Items directly to stdout which is visible in the OMS Agent for Linux log file.

In the OMS general agent configuration file at /etc/opt/microsoft/omsagent/<workspace id>/conf/omsagent.conf:

Comment out the OMS output plugin by adding a # in front of each line

#<match oms.** docker.**>
#  type out_oms
#  log_level info
#  num_threads 5
#  buffer_chunk_limit 5m
#  buffer_type file
#  buffer_path /var/opt/microsoft/omsagent/<workspace id>/state/out_oms*.buffer
#  buffer_queue_limit 10
#  flush_interval 20s
#  retry_limit 6
#  retry_wait 30s
#  max_retry_wait 30m
#</match>

Below the output plugin, uncomment the following section by removing the # in front of each line

<match **>
  type stdout
</match>

My forwarded Syslog messages are not showing up!

Probable Causes

  • The configuration applied to the Linux server does not allow collection of the sent facilities and/or log levels
  • Syslog is not being forwarded correctly to the Linux server
  • The number of messages being forwarded per second are too great for the base configuration of the OMS Agent for Linux to handle

Resolutions

  • Check that the configuration in the OMS Portal for Syslog has all the facilities and the correct log levels
    • OMS Portal > Settings > Data > Syslog, check also if checkbox Apply below configuration to my machines is enabled
  • Check that native syslog messaging daemons (rsyslog, syslog-ng) are able to recieve the forwarded messages
  • Check firewall settings on the Syslog server to ensure that messages are not being blocked
  • Simulate a Syslog message to OMS using logger command
    • logger -p local0.err "This is my test message"

Why OMS Ruby is shipping OpenSSL?

  • OMS doesn't package OpenSSL, and if you are concerned about the openssl.so file located here /opt/microsoft/omsagent/ruby/lib/ruby/2.6.0/x86_64-linux/, there is nothing to worry about, that file is only the library that Ruby is using the handle most abstractions around calling the OpenSSL native API.
  • Why when I use the 'strings' command I can see the presence of OpenSSL 1.0.1 in openssl.so:
strings /opt/microsoft/omsagent/ruby/lib/ruby/2.6.0/x86_64-linux/openssl.so | grep OpenSSL
OpenSSL H
OpenSSL H
OpenSSL
OpenSSL 1.0.1 14 Mar 2012
  • Well, the 'strings' will only show you the strings used inside a binary, but you can't prove if that binary does include OpenSSL. To prove that, you have to use 'ldd' command which will list all dynamic libraries. From the output below you can see that openssl.so point to libssl.so.1.1 which is the real OpenSSL:
ldd /opt/microsoft/omsagent/ruby/lib/ruby/2.6.0/x86_64-linux/openssl.so
	linux-vdso.so.1 (0x00007ffcca6cf000)
	libssl.so.1.1 => /usr/lib/x86_64-linux-gnu/libssl.so.1.1 (0x00007fc87f14e000)
	libcrypto.so.1.1 => /usr/lib/x86_64-linux-gnu/libcrypto.so.1.1 (0x00007fc87ec83000)
	libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fc87e8e5000)
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fc87e4f4000)
	libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fc87e2d5000)
	libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fc87e0d1000)
	/lib64/ld-linux-x86-64.so.2 (0x00007fc87f62c000)

I'm unable to connect through my proxy to OMS

Probable Causes

  • The proxy specified during onboarding was incorrect
  • The OMS Service Endpoints are not whitelistested in your datacenter

Resolutions

  • Re-onboard to the OMS Service with the OMS Agent for Linux using the following command with the option -v enabled. This allows verbose output of the agent connecting through the proxy to the OMS Service.
    • sudo /opt/microsoft/omsagent/bin/omsadmin.sh -w <OMS Workspace ID> -s <OMS Workspace Key> -p <Proxy Conf> -v
    • Review documentation for OMS Proxy located here
  • Double check that the following OMS Service endpoints are whitelisted
Agent Resource Ports
*.ods.opinsights.azure.com Port 443
*oms*.azure.com Port 443
*.oms.opinsights.azure.com Port 443
ods.systemcenteradvisor.com Port 443
*.blob.core.windows.net/ Port 443

Advanced Steps for OMS Gateway debugging:

  • Make sure to choose an appropriate port number which is not already used, otherwise you will not be able to reach your proxy. But instead you will get a 400 Bad request HTTP response from a different web service.

  • You can quickly test the connectivity through OMS Gateway, by running the following cURL command curl -x <PROXY_IP>:<PORT> -L https://<WORKSPACE_ID>.oms.opinsights.azure.com -v

Debugging with Logs

By running the previous curl command we get results shown below. By analysing the results, you should see this first message "Proxy replied OK to CONNECT request" which means that the connection was successfully established with the proxy, then our machine and the server exchanged some handshake messages which were followed by a "403 Forbidden" HTTP. Since we didn't provide the correct certificate, the target server responded with 403 code, which is fine for our case and enough to test that we have reached the server endpoint.

$> curl -x <PROXY_IP>:<PORT> -L https://<WORKSPACE_ID>.oms.opinsights.azure.com -v

* Trying <PROXY_IP>...
* Connected to <PROXY_IP> (<PROXY_IP>) port <PORT> (#0)
...
< HTTP/1.1 200 Connection established
* Proxy replied OK to CONNECT request <--- 
* successfully set certificate verify locations:
* SSLv3, TLS handshake, Client hello (1):
* SSLv3, TLS handshake, Server hello (2):
* SSLv3, TLS handshake, Certificate (11):
...
< HTTP/1.1 403 Forbidden
< Content-Type: text/html
* Server Microsoft-IIS/10.0 is not blacklisted
< Server: Microsoft-IIS/10.0
< X-Powered-By: ASP.NET

We can debug the previous curl command by enabling verbose logging on the proxy (OMS Gateway), by simply setting the registry key "LogLevel" to DEBUG under the following path "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\ForwarderService\Configurations". After you restart OMS Gateway, inspect the logs from "Event Viewer > Applications and Services Logs > OMS Gateway Log" and you will see a similar output as below:

Event ID, Task, Category
103, [4],  DEBUG GatewayLogic - Got request from <PROXY_IP>:50368: CONNECT <WORKSPACE_ID>.oms.opinsights.azure.com:443 with headers [Host, <WORKSPACE_ID>.oms.opinsights.azure.com:443];[User-Agent, curl/7.37.0];[Proxy-Connection, Keep-Alive]
200, [4],  DEBUG GatewayServer - Worker thread created to handle the request
200, [16], DEBUG TlsParser - Client SNI: <WORKSPACE_ID>.oms.opinsights.azure.com
200, [16], DEBUG TcpConnection - Client SNI is verified
200, [16], DEBUG TlsParser - [TLS][Handshake] >ClientHello
200, [15], DEBUG GatewayLogic - Connected to server <WORKSPACE_ID>.oms.opinsights.azure.com:443
200, [16], DEBUG GatewayLogic - Start tunneling data ...
200, [15], DEBUG TlsParser - [TLS][Handshake] <CertificateRequest
200, [15], DEBUG TlsParser - [TLS][Handshake] <ServerKeyExchange
200, [15], DEBUG TcpConnection - Server-side passed verification
107, [23], DEBUG TcpConnection - TLS session is successfully verified
200, [23], DEBUG GatewayServer - Worker thread ended
200, [23], DEBUG GatewayLogic - End tunneling data
200, [23], DEBUG TcpConnection - Client socket disconnected

Debugging with Traces

On your servers, you can capture network traffic using the tcpdump tool and forward the output to a file. Then you can use Wireshark to analyse your traces.

$> sudo tcpdump -i any -w linux_capture.pcap

By tracing the previous curl command we get the following results on Wireshark.

Please refer to official Wireshark documentation for more details about using the tool.

I'm getting a 403 when I'm trying to onboard!

Probable Causes

  • Date and Time is incorrect on Linux Server
  • Workspace ID and Workspace Key used are not correct

Resolution

  • Check the time on your Linux server with the command date. if the data is +/- 15 minutes from current time then onboarding fails. To correct this update, the date and/or timezone of your Linux server.
  • New! The latest version of the OMS Agent for Linux now notifies you if the time skew is causing the onboarding failure
  • Re-onboard using correct Workspace ID and Workspace Key instructions

I'm seeing a 500 Error and 404 Error in the log file right after onboarding

This is a known issue an occurs on first upload of Linux data into an OMS workspace. This does not affect data being sent or service experience.

I'm getting Errno address already in use in omsagent log file

If you see [error]: unexpected error error_class=Errno::EADDRINUSE error=#<Errno::EADDRINUSE: Address already in use - bind(2) for "127.0.0.1" port 25224> in omsagent.log, it would mean that Linux Diagnostic extension (LAD) is installed side by side with OMS linux extension, and it is using same port for syslog data collection as omsagent.

  • As root execute the following commands (note that 25224 is an example and it is possible that in your environment you see a different port number used by LAD):
/opt/microsoft/omsagent/bin/configure_syslog.sh configure LAD 25229

sed -i -e 's/25224/25229/' /etc/opt/microsoft/omsagent/LAD/conf/omsagent.d/syslog.conf

The user will then have to edit the correct rsyslogd or syslog_ng config file and change the LAD-related configuration to write to port 25229.

  • If the VM is running rsyslogd, the file to be modified is /etc/rsyslog.d/95-omsagent.conf (if it exists, else /etc/rsyslog)
  • If the VM is running syslog_ng, the file to be modified is /etc/syslog-ng/syslog-ng.conf
  • Restart omsagent sudo /opt/microsoft/omsagent/bin/service_control restart
  • Restart syslog service

I'm unable to uninstall omsagent using purge option

Probable Causes

  • Linux diagnostic extension is installed
  • Linux diagnostic extension was installed and uninstalled but you still see an error about omsagent being used by mdsd and can not be removed.

Resolution

  • Uninstall LAD extension.
  • Remove LAD files from the machine if they present in the following location: /var/lib/waagent/Microsoft.Azure.Diagnostics.LinuxDiagnostic-<version>/ and /var/opt/microsoft/omsagent/LAD/

Restoring and fixing current agent install:

Follow these steps to recover from a bad state, like changing/removing important files under /[etc/var]/opt/microsoft/omsagent:

  • Extension Install:
    • Purge the agent manually: wget https://raw.githubusercontent.com/Microsoft/OMS-Agent-for-Linux/master/installer/scripts/onboard_agent.sh && sh onboard_agent.sh --purge
    • Got to Azure Portal and remove the OMS extension, wait until you successfully remove the extension.
    • From Azure Portal, install OMS extension again, and you should be able to see the agent successfully installed.
  • Github Install:
    • Download from Github the same agent version that you have installed, and run upgrade command with --force: sh omsagent*.sh --upgrade --force -w <WORKSPACE_ID> -s <SECRET>.

My Nagios data is not showing up in the OMS Portal!

Probable Causes

  • omsagent user does not have permissions to read from Nagios log file
  • Nagios source and filter have not been uncommented from omsagent.conf file

Resolutions

  • Add omsagent user to read from Nagios file instructions
  • In the OMS Agent for Linux general configuration file at /etc/opt/microsoft/omsagent/<workspace id>/conf/omsagent.conf ensure that both the Nagios source and filter are uncommented
<source>
  type tail
  path /var/log/nagios/nagios.log
  format none
  tag oms.nagios
</source>

<filter oms.nagios>
  type filter_nagios_log
</filter>

I'm not seeing any Linux data in the OMS Portal

Probable Causes

  • Onboarding to the OMS Service failed
  • Connection to the OMS Service is blocked
  • VM was rebooted
  • OMI package was manually upgraded to a newer version compared to what was installed by OMS Agent package
  • OMI is frozen, blocking OMS agent
  • DSC resource logs "class not found" error in omsconfig.log log file
  • OMS Agent for Linux data is backed up
  • DSC logs "Current configuration does not exist. Execute Start-DscConfiguration command with -Path parameter to specify a configuration file and create a current configuration first." in omsconfig.log log file, but no log message exists about PerformRequiredConfigurationChecks operations.

Resolutions

  • Install all dependencies like auditd package
  • Check if onboarding the OMS Service was successful by checking if the following file exists: /etc/opt/microsoft/omsagent/<workspace id>/conf/omsadmin.conf
  • Re-onboard using the omsadmin.sh command line instructions
  • If using a proxy, check proxy troubleshooting steps above
  • In some Azure distribution systems omid OMI server daemon does not start after Virtual machine is rebooted. This will result in not seeing Audit, ChangeTracking or UpdateManagement solution related data. Workaround is manually start omi server by running sudo /opt/omi/bin/service_control restart
  • After OMI package is manually upgraded to a newer version it has to be manually restarted for OMS Agent to conitnue functioning. This step is required for some distros where OMI server does not automatically start after upgrade. Please run sudo /opt/omi/bin/service_control restart to restart OMI.
  • In some situations, OMI can become frozen. The OMS agent may enter a blocked state waiting for OMI, blocking all data collection. The OMS agent process will be running but there will be no activity, evidenced by no new log lines (such as sent heartbeats) present in omsagent.log. Restart OMI with sudo /opt/omi/bin/service_control restart to recover the agent.
  • If you see DSC resource "class not found" error in omsconfig.log, please run sudo /opt/omi/bin/service_control restart
  • In some cases, when the OMS Agent for Linux cannot talk to the OMS Service, data on the Agent is backed up to the full buffer size: 50 MB. The OMS Agent for Linux should be restarted by running the following command /opt/microsoft/omsagent/bin/service_control restart
  • Note: This issue is fixed in Agent version >= 1.1.0-28
  • If omsconfig.log log file does not indicate that PerformRequiredConfigurationChecks operations are running periodically on the system, there might be a problem with the cron job/service. Make sure cron job exists under /etc/cron.d/OMSConsistencyInvoker. If needed run the following commands to create the cron job:
mkdir -p /etc/cron.d/
echo "*/15 * * * * omsagent /opt/omi/bin/OMSConsistencyInvoker >/dev/null 2>&1" | sudo tee /etc/cron.d/OMSConsistencyInvoker

Also, make sure the cron service is running. You can use service cron status (Debian, Ubuntu, SUSE) or service crond status (RHEL, CentOS, Oracle Linux) to check the status of this service. If the service does not exist, you can install the binaries and start the service using the following:

On Ubuntu/Debian:
# To Install the service binaries
sudo apt-get install -y cron
# To start the service
sudo service cron start
On SUSE:
# To Install the service binaries
sudo zypper in cron -y
# To start the service
sudo systemctl enable cron
sudo systemctl start cron
On RHEL/CentOS:
# To Install the service binaries
sudo yum install -y crond
# To start the service
sudo service crond start
On Oracle Linux:
# To Install the service binaries
sudo yum install -y cronie
# To start the service
sudo service crond start

My portal side configuration for (Syslog/Linux Performance Counter) is not being applied

Probable Causes

  • The OMS Agent for Linux Configuration Agent has not picked up the latest portal side configuration
  • The changed settings in the portal were not applied

Resolutions

Background: omsconfig is the OMS Agent for Linux configuration agent that looks for new portal side configuration every 5 minutes. This configuration is then applied to the OMS Agent for Linux configuration files located at /etc/opt/microsoft/omsagent/conf/omsagent.conf.

  • In some cases the OMS Agent for Linux configuration agent might not be able to communicate with the portal configuration service resulting in latest configuration not being applied.

  • Check that the omsconfig agent is installed

  • dpkg --list omsconfig or rpm -qi omsconfig

  • If not installed, reinstall the latest version of the OMS Agent for Linux

  • Check that the omsconfig agent can communicate with the OMS Portal Service

    • Run the following command sudo su omsagent -c 'python /opt/microsoft/omsconfig/Scripts/GetDscConfiguration.py'
    • This command returns the Configuration that agent sees from the portal including Syslog settings, Linux Performance Counters, and Custom Logs
    • If this command fails run the following command sudo su omsagent -c 'python /opt/microsoft/omsconfig/Scripts/PerformRequiredConfigurationChecks.py'. This command forces the omsconfig agent to talk to the OMS Portal Service and retrieve latest configuration.

I'm Not Seeing My Custom Log Data in the OMS Portal

Probable Causes

  • Onboarding to OMS Service failed
  • The setting "Apply the following configuration to my Linux Servers" has not been check marked
  • omsconfig has not picked up the latest Custom Log from the portal
  • OMS Agent for Linux user omsagent is unable to access the Custom Log due to permissions or path not being found. In that case osmagent.log will contain one of these messages:
    • [DATETIME] [info]: <path> not found. Continuing without tailing it.
    • [DATETIME] [info]: <path> does not exist or not accessible. Cannot tail the file. Skipping.
    • [DATETIME] [info]: <path> is excluded since it's unreadable or doesn't have proper permissions.

Resolutions

  • Check if onboarding the OMS Service was successful by checking if the following file exists: /etc/opt/microsoft/omsagent/<workspace id>/conf/omsadmin.conf

  • Re-onboard using the omsadmin.sh command line instructions

  • In the OMS Portal under Settings ensure that the following checkbox is checked

  • Check that the omsconfig agent can communicate with the OMS Portal Service

    • Run the following command sudo su omsagent -c 'python /opt/microsoft/omsconfig/Scripts/GetDscConfiguration.py'
    • This command returns the Configuration that agent sees from the portal including Syslog settings, Linux Performance Counters, and Custom Logs
    • If this command fails run the following command sudo su omsagent -c 'python /opt/microsoft/omsconfig/Scripts/PerformRequiredConfigurationChecks.py'. This command forces the omsconfig agent to talk to the OMS Portal Service and retrieve latest configuration.

Background: Instead of the OMS Agent for Linux user running as a privileged user, root - The OMS Agent for Linux runs as the omsagent user. In most cases explicit permission must be granted to this user in order for certain files to be read. For custom log collection, read permission is required. To grant permission to omsagent user run the following commands:

  • Add the omsagent user to specific group sudo usermod -a -G <GROUPNAME> <USERNAME>

  • Grant owner read access to the required file sudo chmod -R ugo+rx <FILE DIRECTORY>

  • Restart the omsagent service: sudo /opt/microsoft/omsagent/bin/service_control restart

  • There is a known issue with a Race Condition in OMS Agent for Linux version <1.1.0-217. After updating to the latest agent run the following command to get the latest version of the output plugin

  • sudo cp /etc/opt/microsoft/omsagent/sysconf/omsagent.conf /etc/opt/microsoft/omsagent/<workspace id>/conf/omsagent.conf

I'm re-onboarding to a new workspace

When you try to re-onboard an agent to a new workspace, OMS Agent configuration needs to be cleaned up before re-onboarding. To clean up old configuration from the agent, run the shell bundle with --purge

sudo sh ./omsagent-*.universal.x64.sh --purge

Or

sudo sh ./onboard_agent.sh --purge

You can continue re-onboarding after using the --purge option

OMS extension on Azure portal is marked as failed state: Provisioning failed

Probable Causes

  • OMS Agent has been removed on OS side
  • OMS Agent service is down/disabled or not configured

Resolution (step by step)

  • Remove extension from Azure portal
  • Install omsagent with instructions
  • On-boarding:
 sudo /opt/microsoft/omsagent/bin/omsadmin.sh -w WORKSPACE_ID -s KEY -v
  • Restart omsagent sudo /opt/microsoft/omsagent/bin/service_control restart
  • Wait for couple of hours, provisioning state would change to Provisioning succeeded

OMS Agent upgrade on-demand

Probable Causes

  • OMS packages on the host are outdated

Resolution (step by step)

  • Check the latest relase on page
  • Download install script (1.4.2-124 as example version) :
wget https://github.com/Microsoft/OMS-Agent-for-Linux/releases/download/OMSAgent_GA_v1.4.2-124/omsagent-1.4.2-124.universal.x64.sh
  • Upgrade packages by executing sudo sh ./omsagent-*.universal.x64.sh --upgrade

I see omiagent using 100% CPU

Probable Causes

A regression in nss-pem package v1.0.3-5.el7 caused a severe performance issue, that we've been seeing come up a lot in Redhat/Centos 7.x distributions.

To learn more about this issue, check the following documentation:

  • Bug 1667121 performance regression in libcurl caused by the use of PK11_CreateManagedGenericObject() [rhel-7.6.z]

Alternatively, if running in a Docker container, using devicemapper will often cause OMI to use 100% CPU.

  • You can check if devicemapper is being used by running:
docker info | grep 'Storage Driver'
  • Unfortunately we don't support devicemapper, as Docker has also deprecated it. Resolution would be to uninstall devicemapper and restart OMS containers.

Debugging omiagent high CPU

Performance related bugs don't happen all the time, and they are very difficult to reproduce. If you experience such issue with omiagent you should use the script omiHighCPUDiagnostics.sh which will collect the stack trace of the omiagent when exceeding a certain threshold.

  • Download the script
wget https://raw.githubusercontent.com/microsoft/OMS-Agent-for-Linux/master/tools/LogCollector/source/omiHighCPUDiagnostics.sh
  • Run diagnostics for 24 hours with 30% CPU threshold
bash omiHighCPUDiagnostics.sh --runtime-in-min 1440 --cpu-threshold 30
  • Callstack will be dumped in omiagent_trace file, If you notice many Curl and NSS function calls as below example, follow resolution steps in the following section.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
0x00007fbfc7bd6736 in pem_FindObjectsInit () from /lib64/libnsspem.so
#0  0x00007fbfc7bd6736 in pem_FindObjectsInit () from /lib64/libnsspem.so
#1  0x00007fbfc7bdddb8 in nssCKFWSession_FindObjectsInit () from /lib64/libnsspem.so
#2  0x00007fbfc7be2f19 in NSSCKFWC_FindObjectsInit () from /lib64/libnsspem.so
#3  0x00007fbfd795f85f in find_objects () from /lib64/libnss3.so
#4  0x00007fbfd795fe08 in nssToken_FindObjectsByTemplate () from /lib64/libnss3.so
#5  0x00007fbfd7960e19 in nssToken_FindTrustForCertificate () from /lib64/libnss3.so
#6  0x00007fbfd79599c3 in nssTrustDomain_FindTrustForCertificate () from /lib64/libnss3.so
#7  0x00007fbfd795dd4c in nssTrust_GetCERTCertTrustForCert () from /lib64/libnss3.so
#8  0x00007fbfd795e0f8 in stan_GetCERTCertificate () from /lib64/libnss3.so
#9  0x00007fbfd7957e7d in nssCertificate_GetDecoding () from /lib64/libnss3.so
#10 0x00007fbfd795ce38 in nssCertificateArray_FindBestCertificate () from /lib64/libnss3.so
#11 0x00007fbfd7920f48 in PK11_FindCertFromNickname () from /lib64/libnss3.so
#12 0x00007fbfd8d186dc in nss_load_cert () from /opt/omi/lib/libcurl.so.3
#13 0x00007fbfd8d5186c in nss_connect_common () from /opt/omi/lib/libcurl.so.3
#14 0x00007fbfd8d47aee in Curl_ssl_connect_nonblocking () from /opt/omi/lib/libcurl.so.3
#15 0x00007fbfd8d1eddd in Curl_http_connect () from /opt/omi/lib/libcurl.so.3
#16 0x00007fbfd8d2e735 in Curl_protocol_connect () from /opt/omi/lib/libcurl.so.3
#17 0x00007fbfd8d41e5b in multi_runsingle () from /opt/omi/lib/libcurl.so.3
#18 0x00007fbfd8d423b1 in curl_multi_perform () from /opt/omi/lib/libcurl.so.3
#19 0x00007fbfd8d39623 in curl_easy_perform () from /opt/omi/lib/libcurl.so.3
#20 0x00007fbfd9281b23 in Pull_SendStatusReport (lcmContext=<optimized out>, metaConfig=<optimized out>, statusReport=0x7fbf83da7450, isStatusReport=<optimized out>, getActionStatusCode=<optimized out>, extendedError=0x7fbfd5bafbe0) at WebPullClient.c:2919
#21 0x00007fbfd926964c in ReportStatusToServer (lcmContext=0x0, errorMessage=0x0, errorSource=0x0, resourceId=0x0, errorCode=0, bLastReport=0 '\000', isStatusReport=1, instanceMIError=0x0) at LocalConfigManagerHelper.c:4817
#22 0x00007fbfd926a28b in SetLCMStatusBusy () at LocalConfigManagerHelper.c:6733
#23 0x00007fbfd9273a25 in Invoke_TestConfiguration_Internal (param=0x1760410) at LocalConfigurationManager.c:2330
#24 0x00007fbfd9294b97 in _Wrapper (param=0x7fbfd00008c0) at thread.c:33
#25 0x00007fbfdbe25dd5 in start_thread () from /lib64/libpthread.so.0
#26 0x00007fbfdb068ead in clone () from /lib64/libc.so.6

Resolution (step by step)

sudo yum upgrade nss-pem
  • If nss-pem is not available for upgrade (mostly happens on Centos), then downgrade curl to 7.29.0-46.
  • If by mistake you run "yum update", then curl will be upgraded to 7.29.0-51 and the issue will happen again.
sudo yum downgrade curl libcurl
  • Restart OMI:
sudo scxadmin -restart