I.e. don't wait for the complete initial sync first.

contracts.EntitiyFactoryFunc.WithInit() checked for
contracts.Initer every time.
Now it is only done once in common.NewSyncSubject().

This also requires explicit handling of custom variables as we need
to multiplex the original values to handle flat custom variables.

….com/goccy/go-yaml

... not to have GPLv2<->Apache 2.0 (app<->deps) license conflicts.

Make {NotificationHistory,StateHistory,History*}#Id UUID -> SHA1

With this change Icinga DB will insert the environment after each
heartbeat takeover if it does not already exist in the database as
the environment may have changed, although this is likely to happen
very rarely,

Instead of checking whether the environment already exists,
uses an INSERT statement that does nothing if it does.

The default environment of Icinga is the empty string.
In Golang, the zero value of string is also the empty string.
But it makes sense to distinguish whether the name is not set
or set to the empty string. That is possible with this change.

If the environment changes during runtime, we have to restart HA
in order to stop a possibly running config sync and start a new
one.

Previously, we selected each entity from the database.
Now we only select entities that belong to the current environment.

Icinga/icinga2#9036 introduced a new environment ID for
Icinga DB that's written to the icinga:stats stream as field
"icingadb_environment". This commit updates the code to make use of this ID
instead of the one derived from the Icinga 2 Environment constant.

There's a small risk that when the environment ID changes, Icinga DB could
update write into the wrong environment in the database. Therefore,
Icinga/icinga2#9036 introduced a new default
environment ID based on the CA public key so that there should be no cases
where it's required to change the actual environment ID. So if this happens
nonetheless, just bail out.

Fix flawed config keys and validation

Better handling of multiple environments

If a heartbeat is pending,
we log it every 60 seconds anyway.

This info message just pollutes the logs and
for debugging we log the execution anyway.

This info message just pollutes the logs and
for debugging we log the execution anyway

These log messages are not relevant for the info level.

Replace Apache 2.0 licensed gopkg.in/yaml.v3 with MIT licensed github.com/goccy/go-yaml

Signed-off-by: Eric Lippmann <eric.lippmann@icinga.com>

dc7511c introduced a parent context check to suppress logging if
it was canceled.
If it was not canceled, the check overwrites the variable err,
which resets every previous real error to nil,
which then leads to fatals because the OnError callback or
the IsRetryable function expect a non-nil error.
Since I cannot reproduce which logs should have been suppressed by
the changes, I removed them completely.

so they don't interfer.

Co-authored-by: Eric Lippmann <eric.lippmann@icinga.com>

Co-authored-by: Eric Lippmann <eric.lippmann@icinga.com>

Co-authored-by: Eric Lippmann <eric.lippmann@icinga.com>

RuntimeUpdates#Sync(): force FIFO for config updates

Fix broken retry mechanics

…ient

Make NotificationRecipient struct readable

The default Redis PoolSize is 10 * CPU, which isn't enough to run Icinga DB on machines with only 1 CPU.
Most of our open connections come from blocking XREADs. (e.g. Heartbeat, Runtime, Overdue, History)

and don't just crash.

refs #427

Enable is more self-documenting.

When embedding TLS, the field name is implicitly specified as TLS,
which conflicts with the YAML key tls,
which currently does not allow TLS to be enabled.
I'm not entirely sure if this is a bug in goccy/go-yaml,
but if a struct field name other than TLS is specified,
the configuration works again.

…-mysql-connection-427

HA#realize(): re-try transaction on temporary error

refs #437

refs #136

refs #136

refs #136

Rationale: see https://dba.stackexchange.com/q/73136

refs #136

refs #136

refs #136

refs #136

refs #136

By the way avoid duplicate rows in the same upsert chunk to avoid Postgres
error 21000 (ON CONFLICT DO UPDATE command cannot affect row a second time).

refs #136

refs #136

refs #136

refs #136

refs #136

refs #136

PostgreSQL does not allow null bytes in varchar, char and
text fields.

Co-authored-by: Eric Lippmann <eric.lippmann@icinga.com>

Co-authored-by: Eric Lippmann <eric.lippmann@icinga.com>

Co-authored-by: Eric Lippmann <eric.lippmann@icinga.com>

Co-authored-by: Eric Lippmann <eric.lippmann@icinga.com>