How to extract Tasy data without slowing down the production database

A query that runs in two seconds in testing can slow billing down at ten in the morning. What changes when extraction is done right, and the checklist to agree with the DBA.

Clic Health teamUpdated February 24, 20266 min read

The Tasy database is the same one serving registration, the emergency department, pharmacy and billing. When someone asks to extract Tasy data for a BI, IT's first concern is fair: a badly written query at the wrong time slows the system for the people treating patients. It is possible to pull everything management needs without that effect, as long as extraction follows a few rules.

Where extraction weighs

A SELECT in Oracle does not lock the rows it reads, so the risk is not freezing patient care. The risk is consumption: CPU, disk reads, memory and temporary space competing with normal use. The cases we see most:

Reading the whole table on every run. CONTA_PACIENTE_RESUMO and PROCEDIMENTO_PACIENTE pile up tens of millions of rows in a few years. Reading all of it every hour to refresh a dashboard is the fastest way to make the database heavy.
A function on the filtered column. WHERE TRUNC(dt_entrada) = TRUNC(SYSDATE) prevents the date index from being used. The equivalent WHERE dt_entrada >= TRUNC(SYSDATE) AND dt_entrada < TRUNC(SYSDATE) + 1 uses it.
Large joins inside Tasy. Joining account, summary, procedure, payer and unit in one aggregated query forces Oracle to build the whole result before returning the first row. Bringing the tables over and joining at the destination is usually lighter on the hospital.
Long queries during working hours. A 40-minute query at ten in the morning competes with billing closing accounts. The same query at five in the morning bothers nobody.
ORA-01555 (snapshot too old). Very long queries may fail when Oracle can no longer rebuild the data as it was when the query started. Splitting the load into smaller slices solves it.

The usual options

Approach	How it works	Load on Tasy	Note
BI querying Tasy directly	Every dashboard opening fires queries at the production database	High and unpredictable	The approach IT complains about most
Exported reports	Someone runs a report and saves it as a spreadsheet	Low, but manual	Delays, versions and typing errors
Read replica or standby	A copy of the database takes the queries	Almost none	Requires licensing and infrastructure; not every hospital has it
Incremental extraction to an intermediate database	A routine brings only what changed and writes it elsewhere	Low and predictable	The model we recommend and that Clic Health uses

When the hospital already has a read replica, it is the best source for extraction. When it does not, well-configured incremental extraction comes close to the same result.

The rules of light extraction

Read-only database user, with SELECT only on the tables and views needed. No Tasy application user, no administrator, no write permission.
Read-only session. Besides the user, the session itself should be opened read-only. If someone mistakenly puts a write command in a query, the database refuses it.
Incremental for transaction tables. Encounters, accounts, procedures and surgeries have DT_ATUALIZACAO. The load brings only rows changed since the last run, with a margin of a few minutes so it does not miss transactions that committed late. The reason is in Tasy incremental loads.
Full loads only for small master tables. Payers, units, cost centers and items can come in full once a day, before dawn.
Frequency by how the information is used. Encounters and accounts every 5 minutes make sense for the emergency department and bed management. Accounting and costing do not change by the hour; once or twice a day is enough.
A limit per run. A cap on rows per run and a maximum time per query keep a filter mistake from becoming a read of years of history in the middle of the day.
Initial load in slices. Five years of history come in month by month, at night and over the weekend, not in a single query.
Periodic re-read of a recent window. Rows deleted in Tasy never show up in an incremental load. Re-reading, for example, the last 7 days every night and the last 3 months once a week keeps the destination equal to the source.
A log of every run. How long it took, how many rows it brought and whether it failed. Without that log, the first sign of trouble is an outdated dashboard.

What to agree with the DBA

The hospital's DBA (or whoever runs the database at the infrastructure provider) has to approve and follow the extraction. Bring to the conversation:

the list of tables and views to be read and the reason for each one;
the user to be created, its permissions, and confirmation that it is read-only;
the frequencies and schedules of the loads, and when the historical load will run;
the expected volume per run after the initial load;
the existing indexes on the date columns used as filters. Do not create indexes on Tasy tables on your own; if one is missing, the decision belongs to the DBA, after hearing the vendor;
the network path: where extraction starts, where it goes and through which port;
how to switch extraction off in an emergency.

With this documented, IT knows exactly what is running on the database and can measure the effect.

How DataClic does it

DataClic is the program that reads Tasy for Clic Health and the other Clic solutions. It runs on a server at the hospital and follows the rules above by default:

opens the Oracle session in read-only mode and refuses users with administrator privileges;
brings transactions incrementally, with a default 15-minute margin over the last watermark;
re-reads 7 days every night and 3 months every Sunday, to reflect deletions and corrections;
caps each run at 5 million rows and each query at 60 minutes;
loads master tables once a day, at 5 a.m.;
does not let the patient's name, national ID, health card, address, phone, e-mail or date of birth leave Tasy; patients travel as a code;
opens no inbound port at the hospital: it only makes outbound HTTPS connections.

Read rules are digitally signed and verified by DataClic before running. A package from another hospital, or one altered on the way, is refused. Details are in security and in Tasy integration and refresh.

Signs that the current extraction is weighing

Complaints about Tasy being slow, always at the same time of day.
BI queries at the top of the database's heaviest sessions.
Dashboards that take minutes to open, because they query Tasy on the spot.
Loads failing with ORA-01555 or running out of temporary space.
Nobody can say how many queries the BI runs on the database per day.

If two or more apply, review the extraction before adding dashboards.

Who writes this

Domínio has worked for more than 20 years with hospitals that run Tasy and wrote the read rules of the Clic Health standard from that experience. If your IT team wants to review an existing extraction, talk to us on WhatsApp or at comercial@dominiocontabil.com.br.

Keep reading

Tasy and dataTasy incremental loadsThe BI shows 1,212 encounters and Tasy shows 1,215. Most of the time, the difference lies in how the incremental load marks what it has already read. The five cases we see and how to fix each one.April 12, 2026 · 6 min read Integration and refreshDataClic reads the Tasy database with a read-only user, brings only what changed and follows the schedule agreed with the hospital. Nothing is written to Tasy.SecurityHealth data is sensitive personal data under Brazil's LGPD. Clic Health was designed so that what identifies patients never leaves the hospital and each hospital sees only its own data.

More in Tasy and data

Tasy and dataReconcile Tasy with the BIOne wrong number in a board meeting costs trust in the entire BI. The routine we use every month to compare the BI with Tasy, from counts to the trial balance, and what to do when it does not match.July 29, 2026 · 6 min read

Tasy and dataCONTA_PACIENTE_RESUMO in the BITasy rebuilds the account summary on every recalculation. Whoever loads CONTA_PACIENTE_RESUMO like a regular table ends up with items that no longer exist. A 90 account that shows up as 120 or 210, and how to fix it.April 18, 2026 · 6 min read