The audit log: who changed what
Every change you, your API keys, our staff or our operations make to your account is recorded, in
the same moment as the change: the change and its record are saved together or not at all, so there
is no change without its record. Settings → Audit log shows them newest first (owners), and
GET /v1/audit-events serves them to a program.
Not recorded yet, and listed here so that a missing row is never a surprise: the credit a top-up
adds to your wallet when the payment settles (the top-up’s request is recorded, as
wallet.top_up_requested), and the account’s pause for credit beginning or ending. Each will be added to this log, and to this page, when it is.
Not here, because nobody asked for it: what a pipeline does by its change policy on its own, such as adding a table it found or applying a schema change. Those are in the pipeline’s own history. What is here is the policy that chose them, each time someone sets it, and each answer someone gives to a change a table waited on.
What a row says
Section titled “What a row says”- Who: a person (by the address they had at the time), an API key (by its name), Reyansh operations (a command run by our operations team, always with a reason), Reyansh staff (by name, always with a reason), or Reyansh itself (a change nobody asked for, such as switching off a channel that kept failing).
- How:
app,api,operator,staff_consoleorsystem. - What: the action (below) and what it was done to, by its name at the time.
- Before and after of each field it changed. A secret is never recorded: a credential’s value, a channel’s URL or signing secret appears as “changed”, never as its value. A change too large to keep in full (over 8 KB) names its fields and is marked truncated.
- The request that made it (
request_id, the id our support asks you for), and for owners signed in as people, the address and browser it came from.
Who can read it
Section titled “Who can read it”Owners, in Settings → Audit log, and an API key (read or full) through the API, for your own security tooling. Addresses and browsers are shown only to an owner signed in as a person: never to an API key, never in a key’s CSV. A member does not read the audit log; a pipeline’s History tab shows the changes people made to it (Show › Changes people made) to everyone who can see the pipeline.
Filters, pages and the CSV
Section titled “Filters, pages and the CSV”GET /v1/audit-events takes actor (a person’s id, a key’s apk_ id), action (a prefix:
pipeline. for every pipeline change), target, pipeline_id, from and to (RFC 3339; at most
366 days apart, 422 range_too_large beyond), and pages with cursor and limit.
GET /v1/audit-events.csv (and Download as CSV in Settings) is the same filtered range as a file,
at most 200,000 rows. The download is itself recorded (audit_log.exported), before any row is
sent. A cell that begins with =, +, -, @, a tab or a carriage return is prefixed with ', so a
spreadsheet never runs a name someone typed as a formula. A download started by a link or a page on
another site is refused (403 cross_site_request): start it from Settings, or call it with a key. If
the download fails part way, it fails as a broken download, never as a shorter file that looks
complete: download it again. Every row saved before the download began is in it exactly once.
Under Download as CSV, an account whose API keys are on is linked to the history export (Platform export), which copies the audit log, without addresses or browsers, with the account’s other history, into a PostgreSQL database you own, once a day.
How long it is kept
Section titled “How long it is kept”Every row is kept for as long as the account exists, and shown 366 days at a time. The address and browser are cleared from each row 365 days after it, the rest of the row kept.
Our staff and operations
Section titled “Our staff and operations”A member of Reyansh staff gives a reason (support, an incident, billing, a security review, and a note) before opening your account in our staff console, and the row shows their name and that reason. A command our operations team runs against your account, such as granting credit, is recorded with the reason it was given.
Every action
Section titled “Every action”Changes made through the app and the API:
| action | when | fields |
|---|---|---|
pipeline.created |
a pipeline was created | source_id, destination_id, tables, schedule |
pipeline.deleted |
a pipeline was deleted | |
pipeline.started |
a pipeline was asked to run | |
pipeline.stopped |
a pipeline was asked to stop | |
pipeline.schedule_changed |
its schedule changed | schedule |
pipeline.discarded_rows_changed |
what happens to rows the destination discards changed | on_discarded_row, on_rejected_value, record_in_destination |
pipeline.notifications_changed |
it was muted or unmuted | muted, until |
pipeline.alert_thresholds_changed |
its alert thresholds changed | grace_minutes, source_log_percent, source_log_bytes, stall_minutes, freshness_minutes |
pipeline.source_cleanup_acknowledged |
a deleted pipeline’s source clean-up was said to be done | |
pipeline.change_policy_changed |
the pipeline’s answers to schema changes were set | rules (each answer that changed, before and after), confirmed (the kinds whose copy-again answer is confirmed) |
pipeline.source_history_accepted |
a person accepted a changed source history, and the pipeline was started | system_id, from_timeline, to_timeline, switch_point, checkpoint |
table.added |
tables were added to a pipeline | tables |
table.removed |
a table was removed | |
table.recopy_requested |
a table was asked to be copied again | delete_rows_not_in_source |
table.verify_requested |
a table’s row counts were asked to be verified, source against destination | |
table.resumed |
a parked table was resumed | |
table.sync_mode_requested |
a change of the mode a table is kept in was asked for | mode (before and after) |
table.sync_mode_request_cancelled |
a waiting change of a table’s mode was cancelled | mode |
table.purge_requested |
a table’s rows marked deleted were asked to be deleted | |
table.purge_request_cancelled |
a waiting purge of a table’s rows marked deleted was cancelled | |
pipeline.sync_mode_default_changed |
the mode a table added later is kept in was set | default (before and after) |
pipeline.resync_requested |
tables, or every table, were asked to be copied again | scope, tables, intent, slot_name_changed, publication_name_changed, repair (the defect, for a repair re-sync an operator asked for; it is not charged) |
pipeline.sync_requested |
Sync now was pressed | effect |
table.change_policy_changed |
one table’s own answers to schema changes were set, or cleared | rules, confirmed |
table.change_decided |
a resume answered the schema changes a table waited on | answers (each change’s id, its kind and the answer given) |
table.column_settings_changed |
how a table’s columns are synced was set (Sync, Block or Hash) | columns (each changed column’s name, its treatment and the answer about its existing values) |
table.filter_changed |
a table’s row filter was set, changed or removed, or a pending one answered | conditions (how many), match, recopy (whether the table is copied again), rows_outside (keep or delete, on a PostgreSQL destination); never a value of the filter |
source.created |
a source was created | name, kind, credential_ref |
destination.created |
a destination was created | name, kind, credential_ref |
destination.hash_salt_revealed |
an owner read the destination’s hash salt (a read that is recorded) | |
destination.hash_salt_set |
an owner set the destination’s hash salt, before any column was hashed with it | (never the salt) |
destination.hash_salt_previewed |
an owner asked the hash a sample value would get | (never the value or its hash) |
destination.edited |
a Snowflake destination’s settings were edited | warehouse, role, user, credential_ref (each before and after, only those that changed; the credential by its reference, never its key) |
source.edited |
a PostgreSQL source’s connection was edited (PATCH /v1/sources/{id}) |
fields (the names of the connection’s fields that changed: host, port, user, password, tls.mode, tls.trust, tls.ca_pem; never a value), password (recorded as changed, never its value) |
source.trust_revoked |
an approved certificate was removed from a PostgreSQL source (POST /v1/sources/{id}/revoke-trust); its pipelines restart and stop at the TLS step until a new approval |
kind (tls_certificate), fingerprint (the certificate’s SHA-256: public, never a secret) |
ssh_key.created |
the account was given an SSH key for reaching a database through a bastion (POST /v1/ssh-keys) |
fingerprint (the key’s SHA256: fingerprint: public, never the private half) |
ssh_key.retired |
an SSH key was retired: its private half deleted, its record kept (POST /v1/ssh-keys/{id}/retire) |
fingerprint |
agent.created |
a proxy agent was created, awaiting enrollment (POST /v1/agents) |
name (never its enrollment token, not even a hint) |
agent.key_approved |
a member approved the key a proxy agent enrolled, naming its fingerprint (POST /v1/agents/{id}/approve) |
fingerprint (the key’s SHA256: fingerprint: public) |
agent.replaced |
a second connection under one proxy agent’s identity replaced the first, which was closed; recorded by the system | address (where it connected from), previous (where the closed connection was from) |
agent.duplicated |
a proxy agent was replaced three times within ten minutes: two copies of it are running, or its key file was copied; recorded by the system, once per burst | replacements |
agent.deleted |
a proxy agent was deleted: its key and token burned, its connection closed, its record kept (DELETE /v1/agents/{id}) |
name |
credential.created |
a credential was stored | shape; value as changed, never valued |
credential.rotated |
a credential was rotated | value as changed, never valued; for a PostgreSQL connection string a pipeline uses, identity_recorded: each destination or source whose database the rotation recorded, that database (a cluster’s id and a database name), and confirmed (true when your confirm asserted it was the same database, which nothing could prove) |
credential.pending_key_generated |
a key pair’s rotation made its pending key; the key in use changes only at the switch, which is a rotation | pending_key as changed, never valued |
connection.tested |
a connection test ran, or a stored destination’s own test (POST /v1/destinations/{id}/test) |
outcome (passed, failed, or unanswered: usable, with a step that got no answer, which is never a pass), tested (connection for the connection test, destination for a destination tested on its own), kind (the source’s for the connection test; the destination’s for its own test, which has no source) |
notification_channel.created |
a notification channel was added | name, kind, the switches, pipelines, triggers; url and signing_secret as changed |
notification_channel.updated |
a channel was changed | name, the switches, enabled, pipelines, triggers; url as changed |
notification_channel.deleted |
a channel was deleted | |
notification_channel.secret_rotated |
a channel’s signing secret was rotated | signing_secret as changed, overlap_hours, drop_previous |
notification_channel.tested |
a test was sent to a channel | trigger |
notification_channel.disabled |
we switched off a channel that answered 410 or kept failing | why |
notification_settings.changed |
the account’s notification email changed | email_to, errors, warnings, recoveries, triggers |
notification_settings.tested |
a test email was sent | trigger |
notification.resent |
a notification was sent again | scope |
me.notifications_changed |
a person changed their own notification email | pipelines, errors, warnings |
session.signed_in |
a person signed in | method (sso for single sign-on; absent for an email link), recovery |
session.switched |
a person switched to this account from another of theirs (which is not named) | recovery (a recovery owner switching into an account that requires single sign-on) |
session.signed_out |
a person signed out | |
wallet.top_up_requested |
a top-up was asked for | amount_minor, currency |
wallet.credit_granted |
our operations granted credit, with the reason | amount_minor, key |
wallet.alerts_changed |
an owner changed the money alerts or the monthly budget | low_balance_days, low_balance_minor, monthly_budget_minor, budget_action |
audit_log.exported |
the audit log was downloaded as CSV | from, to, filter, rows |
account.viewed |
a member of our staff opened this account’s details, with the reason | |
sso.draft_tested |
an owner’s test sign-in finished | outcome (passed or refused), code (the refusal’s reason) |
sso.identity_linked |
a person’s identity at the identity provider was linked to them, at once or after their mailbox confirmed it | issuer, confirmed_by_email |
sso.identity_mismatch |
a sign-in came from a different identity for a linked address, or a linked identity with a new address, and was refused; every owner is told, with whose single sign-on to unlink if it was expected | issuer, linked_to (the people whose links refused it) |
sso.link_confirmation_sent |
a sign-in needed the person’s mailbox to confirm it | owner_invitation |
sso.remembered_identity_changed |
a remembered browser’s sign-in came back as a different person, and was refused | |
sso.link_denied |
a person answered “No, it wasn’t me” to a confirmation code emailed to them; every owner is told | host (the identity provider the confirmation named), owner_invitation |
sso.link_confirmation_expired |
nobody entered a confirmation code within its 15 minutes (no one is told) | |
sso.identity_unlinked |
an owner unlinked a person’s identity at the identity provider; the sessions made through it ended, and the person links again at their next single sign-on (one row per identity) | issuer, sessions_revoked |
sso.mode_changed |
single sign-on was turned off, allowed or required, by an owner (or, with a reason, by our staff when every recovery owner was gone); a switch to required ends the email-link sessions of everyone but the recovery owners; every owner is told how many people that signed out | mode, sessions_revoked, people_signed_out, recovery_sessions_kept |
sso.jit_changed |
an owner changed whether people at the verified domains may join as members on their first single sign-on | jit |
sso.recovery_owners_changed |
an owner changed who may still sign in by email link while single sign-on is required | recovery_owners |
sso.domain_added |
an owner added a domain; it is pending until its DNS record is found | domain |
sso.domain_verified |
an owner’s check found a domain’s DNS record, and the domain is verified (again, when it had lapsed) | domain |
sso.domain_missing |
the daily check did not find a verified domain’s DNS record, the first day in a row; every owner is told | domain |
sso.domain_lapsed |
a verified domain’s DNS record was missing for 14 days in a row, and the domain lapsed: people at it can no longer join or find the account’s sign-in by their address; members already in are unaffected; every owner is told | domain |
sso.domain_removed |
an owner removed a domain | domain |
sso.client_secret_expiring |
the daily check found that the client secret of the live single sign-on connection expires within 30, 7 or 1 days (the expiry date an owner gave when saving it), or has expired: the date, the threshold, the identity provider, its host and the single sign-on mode then; every owner is told | expires_on, threshold, idp, host, mode |
sso.connection_created |
an owner began setting up single sign-on, choosing the protocol and the identity provider; it stays off, and signs nobody in, until a draft is saved, tested and applied | protocol, idp |
sso.draft_saved |
an owner saved the identity provider’s details as a draft; it signs nobody in until it is tested and applied | issuer, client_id, client_secret (recorded as changed, never shown), client_secret_expires_on |
sso.draft_discarded |
an owner discarded the draft; the live version is unchanged | issuer |
sso.version_applied |
an owner made a tested draft the live version; with a new identity provider, everyone signed in through the old one is signed out (the applying owner too, if they were), everyone linked to it links again at their next sign-in, and the browsers it remembered are forgotten | source, issuer, people_relinking, browsers_forgotten, sessions_revoked |
sso.connection_deleted |
an owner deleted the single sign-on connection, which was off, and its versions, domains, linked identities and recovery owners with it |
Read from the account’s two other trails, people and keys (the same rows Settings → Members and API keys show):
| action | when | fields |
|---|---|---|
invitation.created |
someone was invited | role |
invitation.resent |
an invitation was sent again | |
invitation.revoked |
an owner withdrew an invitation | |
invitation.withdrawn |
an invitation was withdrawn because its inviter lost an owner’s access | |
invitation.declined |
an invitation was declined | |
invitation.accepted |
someone joined | role |
member.role_changed |
a person’s role changed | role |
member.removed |
a person was removed | role |
member.left |
a person left | role |
member.joined_by_sso |
someone at a verified domain joined as a member through single sign-on | role |
api_key.created |
an API key was created | scope |
api_key.rolled |
an API key was rolled | scope |
api_key.revoked |
an API key was revoked | scope |
api_key.exposed |
an API key was sent where it must not be, and refused | scope |