FAQ

Common problems

Is it possible to install PTK as a non-root user?

Yes, PTK has no restrictions on the users it can run with.

Is it possible to install a database cluster through a normal user SSH connection

Yes. When deploying a cluster, it is generally recommended that the SSH user linked to the database service in the configuration file be set to a super user with sudo privileges, so that PTK can automatically create a normal user to run the database and configure the user's ulimit parameters. If you cannot provide a super user, you can also create a system user on the database server in advance, and then use the system user to configure the ssh link information in the PTK configuration file to install the database cluster, but you cannot perform the installation of MogHA and other components that require sudo privileges.

Do I Need to Pre-Install Python3?

Yes. Python 3.6 or Python 3.7 needs to be pre-installed because the psutil (python system and process utilities) dynamic library in the OM package supports only this two versions.

PTK and MogDB themselves do not depend on Python3 but gs_om and gs_guc do. Therefore, Python needs to be pre-installed to make sure that the environment meets the requirement.

Can I Install a Database on an Operating System that Is Not Supported by PTK?

Yes.

PTK itself provides a completed and general installation process. It automatically judges the database installation package to be downloaded based on the customer system by default. However, PTK cannot determine the installation package that is really needed when the operating system or operating system version is not supported by PTK. In this case, you can use the -p parameter in the ptk install command to specify an installation package. In view of the fact that the installation is not tested, it may fail. This is normal.

When you try to choose an installation package to adapt to your system, you can comply with the following rules:

Make judgment based on the CPU command set architecture.
```
uname -m
```
For arm64/aarch64, use the openEuler arm64 installation package.

For x86_64, perform the following operation make judgment.
Make judgment based on the glibc version.
```
getconf GNU_LIBC_VERSION
```
If the glibc version is 2.17 or earlier than 2.17, use the centos7 x86_64 installation package. Otherwise, use the openEuler x86_64 installation package.

Is gs_om Still Supported After a Database Is Installed Using PTK?

Yes. PTK-installed clusters are compatible with the gs_om tool, but do not support tools such as gs_upgrade, gs_expansino, gs_dropnode, and other tools that involve changes to the cluster structure.

Note: gs_om needs user SSH mutual-trust between nodes. After the database is installed using PTK, user SSH mutual-trust is generally established. If it is not established, you need to manually perform the establishment.

How Can I Query the MogDB Versions Supported by PTK?

You can run the ptk candidate db command to query the MogDB versions supported by PTK.

How Can I Query the Operating Systems Supported by PTK?

You can run the ptk candidate os command to query the operating systems supported by PTK.

Can I use PTK to install CM?

Yes. CM can be installed automatically by specifying the --install-cm parameter when installing a database cluster.

Does PTK have to be on the same server as the database?

No. The PTK tool is a centralized multi-cluster management tool, so we prefer to recommend that PTK be installed independently on a single management server, with the database installed on a node within a database-specific server cluster.

If I create a new cluster by replicating a virtual machine, how do I change the IP to make PTK work?

For this case, we first need to handle it differently depending on how PTK is used:

1. PTK and clusters are controlled one-to-one

# Let's say there are two clusters, each managed using two PTKs

PTK（1） ---> Cluster1

PTK（2） ---> Cluster2

In this case, you can change the IP address by using PTK(2) to execute the following command:

ptk cluster -n <CLUSTER2> refresh --replace-ip <OLD_IP>=<NEW_IP>

2. PTK centralized control of multiple clusters

     ┌----> Cluster 1
PTK--┤
     └----> Cluster 2

In this case, you can't replace the IP directly, if you execute refresh to replace the IP, it will cause PTK to change from controlling cluster 1 to controlling cluster 2, and ptk ls will not see cluster 1.

In this case, you need to copy the metadata of PTK to create the metadata of cluster 2:

cp -r `ptk env PTK_DATA_DIR`/<CLUSTERNAME1> `ptk env PTK_DATA_DIR`/<CLUSTERNAME2>
vim `ptk env PTK_DATA_DIR`/<CLUSTERNAME2>/topology.yml 
# Modify cluster name and IP information
ptk cluster -n <CLUSTERNAME2> refresh

Common errors

What Can I Do If the Error "cmd init: ssh: setenv failed" Is Reported During Installation?

This error is reported because the sshd service on the target server does not support setting of the LANG environment variable during connection.

You need to perform the following command on the target server to update the sshd configuration file and then restart the sshd service.

sudo echo  "AcceptEnv LANG LC_*" >> /etc/ssh/sshd_config && sudo systemctl restart sshd

What Can I Do If the Database Initialization Error "undefined symbol: EVP_KDF_ctrl" Is Reported?

The installation package is not matched with the host OS. This error typically occurs in the scenario where a CentOS 7 installation package is installed on a system with the high glibc version.

The solution is to replace the installation package with an openEuler package.

What Can I Do If the Error "can not combine with 'abrt-hook-ccpp'" Is Reported?

You can use the abrt(automatic bug report too) service agent to dump the core file. This may lead to core file loss or database downtime.

What is the abrt service

You can fix this bug by modifying the kernel.core_pattern parameter.

# The value can be customized based on the actual requirement.
sysctl -w kernel.core_pattern=/var/log/coredump/%e.%p.%u.%t.core

What Should I Do If the Error `FATAL: semctl(578486307, 7, SETVAL, 0) failed: Invalid argument` Is Reported?

You can check whether the value of RemoveIPC in the /etc/systemd/logind.conf directory is no. If it is, restart the server.

What Should I Do If the Error `fail to new sftp client: EOF` Is Reported?

This error means that PTK failed to create a sftp connection to the target machine. You can manually run sftp -vvv USER@HOST on the machine where PTK is installed to see the error message in the debug log.

This problem is usually caused by a non-existent or inaccessible sftp-server file configured by Subsystem sftp in the /etc/ssh/sshd_config file on the target machine.

FAQ