Troubleshooting

Connect to the database, and run show log_hostname to check whether log_hostname is enabled in the database.

If it is enabled, the database kernel will use DNS to check the name of the host where the client is deployed. If the host where the the host where the CN resides database is configured with an incorrect or unreachable DNS server, the database connection will take a long time to set up. For more details about log_hostname, see the section "GUC parameters".

Problems are difficult to locate in this scenario. Try using the strace Linux trace command.

strace gsql -U MyUserName -d gaussdb -h 127.0.0.1 -p 23508 -r -c '\q'
Password for MyUserName:

The database connection process will be printed on the screen. If the following statement takes a long time to run:

sendto(3, "Q\0\0\0\25SELECT VERSION()\0", 22, MSG_NOSIGNAL, NULL, 0) = 22
poll([{fd=3, events=POLLIN|POLLERR}], 1, -1) = 1 ([{fd=3, revents=POLLIN}])

It can be determined that the database executes the SELECT VERSION() statement slowly.

After the database is connected, you can run the explain performance select version() statement to find the reason why the initialization statement was run slowly. For more information, see "SQL Optimization > Introduction to the SQL Execution Plan" in the Developer Guide.

An uncommon scenario is that the disk of the machine where the database CN resides is full or faulty, affecting queries and leading to user authentication failures. As a result, the connection process is suspended. To solve this problem, simply clear the data disk space of the database CN.

Adapt the steps of troubleshooting slow initialization statement execution. Use strace. If the following statement is run slowly:

connect(3, {sa_family=AF_FILE, path="/home/test/tmp/gaussdb_llt1/.s.PGSQL.61052"}, 110) = 0

Or,

connect(3, {sa_family=AF_INET, sin_port=htons(61052), sin_addr=inet_addr("127.0.0.1")}, 16) = -1 EINPROGRESS (Operation now in progress)

It indicates that the physical connection between the client and the database is set up slowly. In this case, check whether the network is unstable or has high throughput.

This problem occurs generally because an unreachable IP address or port number was specified. Check whether the values of -h and -p parameters are correct.

This problem occurs generally because an incorrect username or password was entered. Contact the database administrator to check whether the username and password are correct.

For security purposes, remote login in trust mode is forbidden. In this case, you need to modify the connection authentication information in the pg_hba.conf file. Contact the administrator

Do not modify the configurations of database cluster hosts in the pg_hba.conf file. Otherwise, the database may become faulty. It is recommended that service applications be deployed outside the database cluster.

Run the SQL statement show unix_socket_directory to check whether the unix socket directory used by the database CN is the same as that specified by the environment variable $PGHOST in the shell directory.

If they are different, set $PGHOST to the directory specified by unix_socket_directory.

This problem occurs because the version of libpq.so used in the environment does not match that of gsql. Run the ldd gsql command to check the version of the loaded libpq.so, and then load correct libpq.so by modifying the environment variable LD_LIBRARY_PATH.

Modify the LD_LIBRARY_PATH environment variable by referring to the example command provided below. In this example, ${path_to_correct_libpq_dir} indicates the directory where libpq.so is located.

export LD_LIBRARY_PATH=${path_to_correct_libpq_dir}:$LD_LIBRARY_PATH

This problem occurs because the version of libpq.so used in the environment does not match that of gsql (or the PostgreSQL libpq.so exists in the environment). Run the ldd gsql command to check the version of the loaded libpq.so, and then load correct libpq.so by modifying the environment variable LD_LIBRARY_PATH.

Is the server running on host "xx.xxx.xxx.xxx" and accepting TCP/IP connections on port xxxx?

This problem is caused by network connection faults. Check the network connection between the client and the database server. If you cannot ping from the client to the database server, the network connection is abnormal. Contact network management personnel for troubleshooting.

ping -c 4 10.10.10.1
PING 10.10.10.1 (10.10.10.1) 56(84) bytes of data.
From 10.10.10.1: icmp_seq=2 Destination Host Unreachable
From 10.10.10.1 icmp_seq=2 Destination Host Unreachable
From 10.10.10.1 icmp_seq=3 Destination Host Unreachable
From 10.10.10.1 icmp_seq=4 Destination Host Unreachable
--- 10.10.10.1 ping statistics ---
4 packets transmitted, 0 received, +4 errors, 100% packet loss, time 2999ms

DETAIL: User does not have CONNECT privilege.

This problem occurs because the user does not have the permission to access the database. To solve this problem, perform the following steps:

1. Connect to the database as the system administrator dbadmin.

   gsql -d gaussdb -U dbadmin -p 8000

2. Grant the user with the permission to access the database.

   GRANT CONNECT ON DATABASE gaussdb TO user1;

   

   Actually, some common misoperations may also cause a database connection failure, for example, entering an incorrect database name, username, or password. In this case, the client tool will display the corresponding error messages.

   gsql -d gaussdb -p 8000
   gsql: FATAL:  database "gaussdb" does not exist
   
   gsql -d gaussdb -U user1 -p 8000
   Password for user user1:
   gsql: FATAL:  Invalid username/password,login denied.

This problem occurs because the number of system connections exceeds the allowed maximum. Contact the DBA database administrator to release unnecessary sessions.

You can check the number of connections as described in Table 1.

You can view the session status in the PG_STAT_ACTIVITY view. To release unnecessary sessions, use the pg_terminate_backend function.

select datid,pid,state from pg_stat_activity;

 datid |       pid       | state  
-------+-----------------+--------
 13205 | 139834762094352 | active
 13205 | 139834759993104 | idle
(2 rows)

The value of pid is the thread ID of the session. Terminate the session using its thread ID.

SELECT PG_TERMINATE_BACKEND(139834759993104);

If a command output similar to the following is displayed, the session is successfully terminated:

PG_TERMINATE_BACKEND
----------------------
 t
(1 row)

SELECT ROLNAME,ROLCONNLIMIT FROM PG_ROLES WHERE ROLNAME='user1';
 rolname | rolconnlimit
---------+--------------
 user1    |           -1
(1 row)

SELECT COUNT(*) FROM dv_sessions WHERE USERNAME='user1';

 count
-------
     1
(1 row)

SELECT DATNAME,DATCONNLIMIT FROM PG_DATABASE WHERE DATNAME='gaussdb';

 datname  | datconnlimit
----------+--------------
 gaussdb |           -1
(1 row)

SELECT COUNT(*) FROM PG_STAT_ACTIVITY WHERE DATNAME='gaussdb';
 count 
-------
     1
(1 row)

SELECT COUNT(*) FROM dv_sessions;
 
 count
-------
     10
(1 row)

When gsql initiates a connection request to the database, a 5-minute timeout period is used. If the database cannot correctly authenticate the client request and client identity within this period, gsql will exit the connection process for the current session, and will report the above error.

Generally, this problem is caused by the incorrect host and port (that is, the xxx part in the error information) specified by the -h and -p parameters. As a result, the communication fails. Occasionally, this problem is caused by network faults. To resolve this problem, check whether the host name and port number of the database are correct.

Check whether CN logs contain information similar to "FATAL: cipher file "/data/coordinator/server.key.cipher" has group or world access". This error is usually caused by incorrect tampering with the permissions for data directories or some key files. For details about how to correct the permissions, see related permissions for files on other normal instances.

In pg_hba.conf of the target CN, the authentication mode is set to gss for authenticating the IP address of the current client. However, this authentication algorithm cannot authenticate clients. Change the authentication algorithm to sha256 and try again. Contact the administrator

• Do not modify the configurations of database cluster hosts in the pg_hba.conf file. Otherwise, the database may become faulty.
• It is recommended that service applications be deployed outside database cluster.

Generally, this problem is caused by changes in loading the shared dynamic library (.so file in Linux) during process running. Alternatively, if the process binary file changes, the execution code for the OS to load machines or the entry for loading a dependent library will change accordingly. In this case, the OS kills the process for protection purposes, generating a core dump file.

To resolve this problem, try again. In addition, do not run service programs in a cluster during O&M operations, such as an upgrade, preventing such a problem caused by file replacement during the upgrade.

A possible stack of the core dump file contains dl_main and its function calling. The file is used by the OS to initialize a process and load the shared dynamic library. If the process has been initialized but the shared dynamic library has not been loaded, the process cannot be considered completely started.