Twitter | Search | |
Franck Pachot
👨🏽‍💻> DBA & Database Evangelist at 👨‍🎓> 🔸 DB 🅾️ 12c 📝> 🎧> 🍻> 🤓 ♠️
22,113
Tweets
948
Following
5,504
Followers
Tweets
Franck Pachot 3h
Yes, I prefer that queries my API than generating crazy queries.
Reply Retweet Like
Franck Pachot retweeted
Jeremy Schneider 15 Oct 19
Oh, think I found the ghost row! pg-11.5 rw root@db1=# select; -- (1 row)
Reply Retweet Like
Franck Pachot retweeted
Jonathan Lewis 16h
Replying to @FranckPachot
pctfree = 30 is a good idea for "random arrival" indexes; pctfree = 1 for "monotonic increasing arrival"; then there are a few outliers where you have to think about the best strategy for damage limitation.
Reply Retweet Like
Franck Pachot 16h
Replying to @RogerMacNicol
I think all is intended and very good to estimate it with dbms_xplan. Just wanted to warn people that there’s always a PCTFREE () and if we decide to rebuild we should know which pctfree we want.
Reply Retweet Like
Franck Pachot 17h
If you rebuild an index, which PCTFREE do you set? REBUILD does not remove free space. It actually add the free space you specify (or the default 10%). Index rebuild can also increase the size of the index:
Reply Retweet Like
Franck Pachot 18h
Replying to @joelgarry
IMO it is more about "backup of data" vs. "backup of database". A dump is a backup of your data. When you restore it you get your data (and metadata) back. But with a backup of database you recover the whole database, including physical layout and latest transactions.
Reply Retweet Like
Franck Pachot 18h
I can't think about it. Just try it and 🤣&😱:
Reply Retweet Like
Franck Pachot 18h
Totally agree. Same laws, different API and specialized datastore. Nobody can bypass the design step or they pay it later.
Reply Retweet Like
Franck Pachot 18h
Yes, that's why I see using "SQL" as a synonym for "RDBMS" a source of confusion. SQL queries and defines a logical layer. Defining tables in SQL doesn't infer that they are not co-located. RDBMS store tables in different physical segments (usually).
Reply Retweet Like
Franck Pachot 18h
Apologies, my goal was not to be pedantic. I've seen mis-use of terms leading to mis-understanding and bad decisions. People think they have to do in NoSQL all opposite of RDBMS. NoSQL is very different from SQL by its API but normalization concepts may apply on DynamoDB
Reply Retweet Like
Franck Pachot 18h
Which SQL concepts could be used for schema design? SQL is just a language, there’s no design concept. Do you mean normalization? The single-table design referenced by is similar. It “scales” by ensuring physical colocation in the same partition as there are no joins
Reply Retweet Like
Franck Pachot 19h
Replying to @westermanndanie
I never realized that you can call non scalar functions without a from clause. That’s really weird.
Reply Retweet Like
Franck Pachot 19h
Why? Only data processing (joins,aggregations,...) will be done in the database service. There’s still a lot to do in Java for the control and presentation. Your point of view is interesting because maybe I missed something.
Reply Retweet Like
Franck Pachot 20h
Well said. I feel the same. I know people the opposite, like just calling when I ask a question be email 😕
Reply Retweet Like
Franck Pachot 20h
A backup that do not allow PITR is not a backup A backup that do not allow PITR is not a backup A backup that do not allow PITR is not a backup (It is just a copy of data)
Reply Retweet Like
Franck Pachot 20h
Replying to @lukaseder @pg_xocolatl
There may even be some performance improvement with order by. Oracle sqlnet do not repeat columns with identical value than previous one. Don’t know for other databases.
Reply Retweet Like
Franck Pachot 21h
Replying to @vlad_mihalcea
Great, one day Java objects, ORMs and RDBMS will all converge: ingest data and changes as JSON objects and query them on relational views with SQL, maybe through REST API
Reply Retweet Like
Franck Pachot 22h
Out-of-row storage for large text in
Reply Retweet Like
Franck Pachot 22h
That’s exactly how I see MongoDB. Rather than telling ORM devs that they can improve their queries by knowing the database that is behind, rub them up the right way by reducing the concept of database to a simple store for objects. Yes, NoSQL is faster than poor ORM mapping 🤷
Reply Retweet Like
Franck Pachot Oct 21
Yes expdp metadata, impdp sqlfile and put to git 👍 (users and dblink may be excluded for security reason but still helpful to keep somewhere)
Reply Retweet Like