Skip to content

Connection test: every step and message

A connection test runs these steps in this order, and shows each one with how it ended. A step that fails shows a message: the messages are listed below, each with its own link, in the words the app uses.

Step Checked
Find the server for the connection
Open the SSH tunnel for the connection
Reach the database server for the connection
Secure the connection for the connection
Sign in for replication for the connection
Read the server for the connection
Check the server’s settings for the connection
Find the published tables (publication) for a pipeline, with its own names
Find the change bookmark (replication slot) for a pipeline, with its own names
Read the tables for a pipeline, with its own names
Check each table can be read by queries for a pipeline, with its own names
Check row-level security for a pipeline, with its own names
Measure the tables for a pipeline, with its own names
Check the replica’s settings for the connection
Check the server keeps what it reports for the connection
Check the source’s history for a pipeline, with its own names
Check the database provider for the connection
Count the servers behind the address for the connection
Check how the tables are read for a pipeline, with its own names
Reach the destination for the connection
Secure the destination’s connection for the connection
Sign in to the destination for the connection
Check the destination’s permissions for the connection
Check the destination’s table for the connection

A source tested on its own does not run the steps checked for a pipeline: those run with the pipeline’s own names, at its check and each time it starts.

  • Passed
  • Passed, with a warning
  • Failed
  • No answer
  • Skipped
  • Not run
  • Not there yet
Step Message
Check the destination’s permissions The destination database is not UTF8, or its connection does not speak UTF8, and a PostgreSQL destination must be a UTF8 database.
Check the source’s history A restore to a point in time starts a new timeline, which stops this pipeline as a failover does. A disk or VM snapshot, or a backup brought back as it was, keeps the timeline. It is detected for certain while your database is behind what we last read. Once it has caught up, it goes unnoticed, and then rows your database wrote soon after it can be missing from your destination or keep older values, for good on a table that is only inserted into. We tell you when your database restarts. After any restore you know of, re-sync.
Find the server The host name was not found within 5 seconds.
Reach the database server Nothing answered within 10 seconds: a firewall dropping the connection is the usual cause.
Code Message
agent_awaiting_approval The agent enrolled a key that nobody has approved yet. Compare its fingerprint with the line the agent printed, then approve it in Settings, Network and security, Proxy agents.
agent_deleted The agent was deleted. Point this source at another agent.
agent_offline The agent is not connected. Its pipelines resume on their own when it connects.
agent_ticket_refused The connection to the agent was refused for a reason on our side. It is retried.
agent_unresponsive The agent is connected but did not open a connection within 10 s.
approval_needed The server’s certificate is not signed by anything this connection trusts.
aurora_reader This is an Aurora reader instance: connect to the cluster’s writer.
aurora_reader_untested This is an Aurora reader, or this user cannot show it is the writer: connect to your cluster’s writer endpoint.
aurora_unconfirmed A function named as one of Aurora’s own exists outside pg_catalog, so the test could not confirm this is not an Aurora reader.
aurora_writer_unconfirmed AWS’s record of the Aurora cluster does not name this instance its writer: connect to the cluster’s writer.
aurora_writer_unverified AWS’s record of the Aurora cluster’s writer could not be read, so this instance was judged by recovery alone.
auth_cleartext_unverified The password crossed a connection whose server is not verified.
auth_md5_deprecated Signed in by MD5, which PostgreSQL 18 deprecates.
auth_method_refused The server’s sign-in method refused this role.
auth_unsupported The server asked for a sign-in method this service does not use.
bad_password The password was refused. The server says the same for a role that does not exist, or whose password has expired.
certificate_changed The server presents a different certificate from the one approved.
certificate_rejected The server’s certificate was refused.
channel_binding_mismatch The sign-in could not be bound to this TLS session: something between us and the server ends the TLS.
channel_binding_stripped Channel binding was removed on the way: something between us and the server rewrites the sign-in.
channel_binding_unavailable Channel binding is required, and the server did not use it.
client_certificate_required The server requires a client certificate.
client_certificate_role_mismatch The client certificate is not this role’s.
column_list_under_full The publication’s column list does not fit the table’s REPLICA IDENTITY FULL.
configuration The connection’s settings could not be used.
connection_lost The connection was lost part way.
database_missing The database does not exist.
destination_blocked The destination refuses the table.
destination_unusable The destination cannot be used as configured.
durability_unreadable This user cannot read how far the server’s log is on disk, which every sync checks.
encoding_not_utf8 The server’s encoding is not UTF-8.
exact A switchover or a failover that lost nothing continues, and we tell you; one that lost what we read stops before reading and asks you to re-sync.
expired The server’s certificate has expired.
failover_without_standby_slots The change bookmark (replication slot) is kept through a failover, but no standby is listed for it to wait for.
foreign_key_between_tables A foreign key joins two of the tables.
fsync_off fsync is off on this server, so a crash can lose changes it reported as written.
generated_unsent_under_full A generated column is not sent by the publication.
handshake_failed The TLS handshake failed.
hba_refused The server’s pg_hba.conf refused this connection.
hba_reject_line A pg_hba.conf line rejects this connection.
host_key_algorithm_unsupported The bastion offers only a host key this service does not verify (RSA or DSA).
host_key_approval_needed No host key is approved for this bastion yet. Compare the one it presents with the bastion’s own, then approve it.
host_key_changed The bastion presents a different host key from the one approved.
hot_standby_feedback_off This standby has hot_standby_feedback off.
identity_unavailable The server’s identity could not be read.
idle_slot_timeout_short An idle change bookmark (replication slot) is dropped in under a day.
ipv4_only The host has only IPv4 addresses, and this service connects from IPv6.
ipv6_only The host has only IPv6 addresses, and this service connects from IPv4.
key_refused The destination refused the key.
key_withheld The publication withholds a key column.
lock_hint PostgreSQL 13 reads each table in one statement, holding its lock for the whole read.
lock_refused This user may not lock a table to read it: grant SELECT on the table itself.
lock_wait A lock held by another session blocked this read.
name_mismatch The certificate is valid, but issued for another name.
no_connect_privilege The role may not connect to this database.
no_free_sender Every WAL sender is in use.
no_free_slot Every change bookmark (replication slot) the server allows is in use.
no_hba_line The server’s pg_hba.conf has no line for this connection.
no_key A table has no primary key.
no_password The server asked for a password and none was given.
no_tables_picked Checked for each table once you pick them.
no_trust_anchors There is no certificate to verify the server against.
no_writable_columns A table has no column the destination can write.
not_a_table A name you picked is not a table, a view for example. A query-based pipeline syncs a view only with the key columns you name for it (key_columns, through the API); this list picks tables.
not_published A table is not in the publication.
not_replication_role The role cannot replicate: it needs the REPLICATION attribute.
permission_denied The destination refused a permission the pipeline needs.
permission_unanswered The permission test gave no answer, so whether the role may create what a pipeline needs is not known. Test again.
pgoutput_not_allowed The pgoutput plugin is not allowed on this server.
plain_primary After a failover or a planned switchover (such as your provider’s maintenance) we ask you whether to continue, showing what we could check; this user cannot read where the new history began. Grant it REPLICATION to have this judged exactly.
plain_replica A failover upstream of this replica, or its own promotion, continues when everything we can check continues. One that lost transactions we had read and then caught up can go unnoticed: grant this user REPLICATION to have it judged exactly.
privilege_refused The server refused a privilege this sign-in needs: CONNECT on the database, or REPLICATION.
provider_untested Not yet verified on this provider: a crash just before a sync is a case we have not proven safe there.
publication_missing The published tables (publication) do not exist yet.
publication_narrowed The publication covers less than the pipeline asks for.
query_based Not used: a pipeline that reads its tables by queries needs no published tables (publication), change bookmark (replication slot) or logical replication settings.
query_version_too_old Reading tables by queries needs PostgreSQL 13 or later.
read_timed_out A read took longer than 5 seconds.
replica_position_unreadable A server behind this address does not report how far it has replayed: use one replica’s own address.
role_cannot_log_in The role may not log in (NOLOGIN).
role_missing The role does not exist.
role_not_granted The role is not granted to this user.
role_refused The server refused the role: it does not exist, or may not log in.
role_unresolved The destination could not resolve the role.
row_secured Row-level security would hide rows from this pipeline.
scan_large Some tables are large, and every sync reads each one whole.
search_path_unpinned The session did not take the search path the pipeline sets; a pooler or proxy may have dropped it.
server_error The server answered with an error.
server_has_no_tls The server does not offer TLS.
server_unproven The server, or something answering in its place, signed in without proving it holds the password, so the connection was dropped: something between us and the server may be standing in for it.
several This address reaches more than one server: every read waits for one that is not behind what we last read.
sign_in_refused The sign-in was refused for a reason the server’s answer does not let us place.
slot_in_use Another session is reading the change bookmark (replication slot).
slot_invalidated The change bookmark (replication slot) was invalidated.
slot_is_synced_copy The change bookmark (replication slot) is a standby’s synced copy, which cannot be read.
slot_missing The change bookmark (replication slot) does not exist yet.
slot_not_logical The change bookmark (replication slot) is not a logical one.
slot_unreserved The change bookmark (replication slot) no longer reserves its log.
slot_wrong_database The change bookmark (replication slot) belongs to another database.
slot_wrong_plugin The change bookmark (replication slot) uses another output plugin.
source_history_unknown This server’s timeline cannot be read, so a failover would not be detected.
ssh_auth_refused The bastion refused our key: the line below is not in the SSH user’s authorized_keys.
ssh_failed The SSH tunnel failed for a reason on our side.
ssh_forwarding_refused The bastion refused to forward to the database: AllowTcpForwarding or the line’s permitopen does not allow it.
ssh_handshake_failed The SSH session to the bastion could not be set up: the port may not be SSH, the bastion offers no algorithm this service uses, or it sent a host key or key-exchange message that could not be read or did not verify.
ssh_target_unreachable The bastion could not connect to the database.
ssh_unreachable The SSH bastion could not be reached, or the session to it ended.
standby_cancels_reads This replica may cancel our reads to replay changes from the primary.
standby_delay_short This standby cancels long reads within 15 minutes.
synchronous_commit_off synchronous_commit is off on this server, so a crash can lose its last changes.
table_missing A table is not on this server.
timed_out This step did not finish within its time limit.
too_many_connections The server has no free connection.
unlogged_on_standby A replica cannot read an unlogged table: read it from the primary.
unmappable A column’s type has no destination type.
unreachable The connection was refused or reset.
unreadable This could not be read.
version_past_end_of_life PostgreSQL 13 is past its end of life: upgrade when you can.
version_too_old This PostgreSQL version is too old for logical replication here.
version_untested This PostgreSQL major is not one our tests run against.
virtual_under_full A virtual generated column under REPLICA IDENTITY FULL makes the source refuse the table’s UPDATE and DELETE.
wal_level_not_logical Your database does not keep the change log a pipeline reads: its wal_level setting is not logical.
wal_sender_timeout_short wal_sender_timeout is shorter than a minute.
weak_mode Connected under a mode that does not verify the server.
write_loop The destination is the source: the pipeline would read its own writes.
xmin_refused This user may not read a table’s rows with their xmin: grant SELECT on the table.