TIL

_posts/2023-12-10-tech-diary.md

@@ -7,6 +7,12 @@ tags: [diary]
 
 Some random thoughts or learnings.
 
+## 2024-12-05 Thu
+
+I always came cross real models. For example today
+<https://2023.splashcon.org/details/iwaco-2023-papers/5/Borrow-checking-Hylo>.
+Stay focused and consistently!
+
 ## 2024-10-06 Sun
 
 I came across this repo <https://github.com/libusual/libusual> when I dig into

_posts/2024-12-04-postgres-vacuum.md

@@ -13,7 +13,6 @@ The purpose of vacuum:
 3. Prevent transaction id wraparound failure
 
    - how to find the max and min XID in postgres
-   - how is vacuum affected by XID?
    - how dd dashboard show low row count after failover
    - how does it set FrozenTransactionId
    - how MVCC prevents vacuum, and only subset of tables or all tables?
@@ -22,9 +21,7 @@ How does vacuum compare XID? See
 [code](https://github.com/postgres/postgres/blob/a3e6c6f929912f928fa405909d17bcbf0c1b03ee/src/backend/access/transam/transam.c#L280)
 Wrapp around happens
 
-## Autovacuum
-
-### When will autovacuum run?
+## When will autovacuum run?
 
 Autovacuum is controlled by two parameters
 
@@ -97,6 +94,21 @@ VACUUM
 
 You see that it skips 1 page due to buffer pins.
 
+## Freezing
+
+[MVCC in PostgreSQL — 8. Freezing](https://postgrespro.com/blog/pgsql/5967948)
+from postgrespro is the best article talking about freezing I can find online.
+It creates a live experiment that I can follow. The important parameters that
+control the freezing are below:
+
+- vacuum_freeze_table_age: the age of the transaction for which vacuuming
+  ignores the visibility map and looks through all the table pages in order to
+  do freezing
+- vacuum_freeze_min_age: the minimum age of the xmin transaction for which a
+  tuple can be frozen
+- autovacuum_freeze_max_age: forced autovacuuming is launched if
+  pg_class.relfrozenxid is older than this.
+
 ### Logs
 
 `log_autovacuum_min_duration` controls the threshold of long-running autovacuum