FRQ: maintenance jobs for hassio data

[AlexMKX]

Hello. Thank you for such great extension.

My usage implies some sensors to be requested every second. As a result - around 600MB data is collected every day. However, this data contains lot of dupes. Nonetheless it runs on ZFS with dedupe and compression, so storage is not a big problem, the HASS performance degrades anyway.

I created a script to remove state dupes and attribute dupes from pg base and later vacuuming. It would be great to have it in the default installation.

-- create mat view with dupes (non-changed states) and calculate old_state to fix later
create materialized view cleanup_dupes as (with d as (SELECT state_id,
                                                    FIRST_VALUE(STATE_ID) OVER (PARTITION BY ENTITY_ID,
                                                        LAST_CHANGED_TS
                                                        ORDER BY last_updated_ts) AS FST,
                                                    last_value(state_id) OVER (PARTITION BY ENTITY_ID,
                                                        LAST_CHANGED_TS
                                                        ORDER BY last_changed_ts) AS LST
                                             FROM PUBLIC.STATES
                                             WHERE to_timestamp(last_changed_ts) < (NOW() - interval '12 hours'))
                                  select *,
                                         CASE when state_id != fst and state_id != lst then true else false end as to_delete
                                  from d) with no data;

create index dupes_state_id on cleanup_dupes (state_id);
create index dupes_index_fst on cleanup_dupes (fst);

-- maintenance 
-- refresh matview with dupes
refresh materialized view cleanup_dupes;
-- fixup old_state.
UPDATE public.states
SET OLD_STATE_ID = cleanup_dupes.FST
FROM cleanup_dupes
WHERE cleanup_dupes.STATE_ID = STATES.STATE_ID
  AND cleanup_dupes.FST != STATES.STATE_ID
  and states.old_state_id is not  NULL;

-- delete duped states
delete from public.states
where exists(select state_id
             from cleanup_dupes
             where states.state_id = cleanup_dupes.state_id
               and cleanup_dupes.to_delete = true);
-- delete unreferences attributes 
delete from public.state_attributes where not exists (
    select attributes_id from public.states where states.attributes_id=state_attributes.attributes_id
) ;
-- free space
vacuum full analyze state_attributes, states;

[expaso]

Hi @AlexMKX ,

Nice Script! Very usefull!

I only see a problem: to incorporate this into the addon, I have to be sure somebody uses this addon to store the homeassistant recorder database itself. Most people do this, but not all.

I don't think I should sit on the users chair and proactively delete data from their tables. What is your view on this?

[expaso]

Closed as being stale.