PostgreSQL Large Objects and space usage (part 3)

A blog post series would not be complete without a final post about vacuumlo.

In the previous post we have seen that the large objects are split in tuples containing 2048 bytes each one, and each chunk behaves in the very same way as regular tuples.

What distinguish large objects?
NOTE: in PostgreSQL, IT IS possible to store a large amount of data along with the table, thanks to the TOAST technology. Read about TOAST here.

Large objects are not inserted in application tables, but are threated in a different way. The application using large objects usually has a table with columns of type OID. When the application creates a new large objects, a new OID number is assigned to it, and this number is inserted into the application table.
Now, a common mistake for people who come from other RDBMS (e.g. Oracle), think that a large object is unlinked automatically when the row that references
it is deleted. It is not, and we need to unlink it explicitly from the application.

Let’s see it with a simple example, starting with an empty pg_largeobject table:

Let’s insert a new LOB and reference it in the table t:

Another one:

If we delete the first one, the chunks of its LOB are still there, valid:

If we want to get the rid of the LOB, we have to unlink it, either explicitly or by using triggers that unlink the LOB when a record in the application table is deleted.
Another way is to use the binary vacuumlo included in PostgreSQL.
It scans the pg_largeobject_metadata and search through the tables that have OID columns to find if there are any references to the LOBs. The LOB that are not referenced, are unlinked.
ATTENTION: this means that if you use ways to reference LOBs other than OID columns, vacuumlo might unlink LOBs that are still needed!

vacuumlo has indeed unlinked the first LOB, but the deleted tuples are not freed until a vacuum is executed:

So vacuumlo does not do any vacuuming on pg_largeobject table.

PostgreSQL Large Objects and space usage (part 2)

In my previous post I showed how large objects use space inside the table pg_largeobject when inserted.

Let’s see something more:

The table had 2 large objects (for a total of 1024 records):

Let’s try to add another random-padded file:

As expected, because a random sequence of characters cannot be compressed, the size increased again by 171 blocks (see my previous post for the explanation)

If you read this nice series of blog posts by Frits Hoogland, you should know about the pageinspect extension and the t_infomask 16-bit mask.

Let’s install it and check the content of the pg_largeobjects pages:

We already know the mathematics, but we love having all the pieces come together 🙂

We know that: The page header is 24 bytes, and that the line pointers use 4 bytes for each tuple.

The first 4 pages have the lower offset to 452 bytes means that we have (452-24)/4 = 107 tuples.

The 5th page (page number 4) has the lower to 360: (360-24)/4=84 tuples.

The remaining pages have the lower to 36: (36-24)/4 = 3 tuples.

Let’s check if we are right:

Now, let’s delete the 1Mb file and check the space again:

The space is still used and the tuples are still there.

However, we can check that the tuples are no longer used by checking the validity of their t_xmax. In fact, according to the documentation, if the XMAX is invalid the row is at the latest version:

[…] a tuple is the latest version of its row iff XMAX is invalid or t_ctid points to itself (in which case, if XMAX is valid, the tuple is either locked or deleted). […]
 (from htup_details.h lines 87-89).
We have to check the infomask against the 12th bit (2048, or 0x0800)
#define HEAP_XMAX_INVALID       0x0800  /* t_xmax invalid/aborted */

Here we go. The large objects are split in compressed chunks that internally behave the same way as regular rows!

If we import another lob we will see that the space is not reused:

Flagging the tuples as reusable is the vacuum’s job:

The normal vacuum does not release the empty space, but it can be reused now:

If we unlink the lob again and we do a vacuum full, the empty space is released:

PostgreSQL Large Objects and space usage (part 1)

PostgreSQL uses a nice, non standard mechanism for big columns called TOAST (hopefully will blog about it in the future) that can be compared to extended data types in Oracle (TOAST rows by the way can be much bigger). But traditional large objects exist and are still used by many customers.

If you are new to large objects in PostgreSQL, read here. For TOAST, read here.

Inside the application tables, the columns for large objects are defined as OIDs that point to data chunks inside the pg_largeobject table.


Because the large objects are created independently from the table columns that reference to it, when you delete a row from the table that points to the large object, the large object itself is not deleted.

Moreover, pg_largeobject stores by design all the large objects that exist in the database.

This makes housekeeping and maintenance of this table crucial for the database administration. (we will see it in a next post)

How is space organized for large objects?

We will see it by examples. Let’s start with an empty database with empty pg_largeobject:

Just one block. Let’s see its file on disk:

First evidence: the file is empty, meaning that the first block is not created physically until there’s some data in the table (like deferred segment creation in Oracle, except that the file exists).

Now, let’s create two files big 1MB for our tests, one zero-padded and another random-padded:

Let’s import the zero-padded one:

The large objects are split in chunks big 2048 bytes each one, hence we have 512 pieces. What about the physical size?

Just 40k! This means that the chunks are compressed (like the TOAST pages). PostgreSQL uses the pglz_compress function, its algorithm is well explained in the source code src/common/pg_lzcompress.c.

What happens when we insert the random-padded file?

The segment increased of much more than 1Mb! precisely, 1441792-40960 = 1400832 bytes. Why?

The large object is splitted again in 512 data chinks big 2048 bytes each, and again, PostgreSQL tries to compress them. But because a random string cannot be compressed, the pieces are still (average) 2048 bytes big.

Now, a database block size is 8192 bytes. If we subtract the size of the bloch header, there is not enough space for 4 chunks of 2048 bytes. Every block will contain just 3 non-compressed chunks.

So, 512 chunks will be distributed over 171 blocks (CEIL(512/3.0)), that gives:

1400832 bytes!

Depending on the compression rate that we can apply to our large objects, we might expect much more or much less space used inside the pg_largeobject table.

Time for an additional RDBMS platform in this blog?

Since its creation (9 years ago), this blog has been almost only Oracle-oriented. But during my career I worked a lot with other RDBMS technologies… SQL Server, MySQL (and forks), Sybase, PostgreSQL, Progres. Some posts in this blog prove it.

The last two years especially, I have worked a lot with PostgreSQL. In the last few months I have seen many friends and technologists increasing their curiosity in this product. So I think that I will, gently, start blogging also about my experiences with PostgreSQL.

Stay tuned if you are interested!

Italian Oracle User Group: from nothing to something

I am pretty sure that every user group has had its own difficulties when starting. Because it happened that we started the ITOUG just a couple of years ago, I think it is worth to tell the story 🙂

I have been told about a team of italian DBAs willing to create a new user group, back in 2013. I decided to join the team because I was starting my journey in the Oracle Community.

We were coming from different cities (Lausanne, Vicenza, Milano, Roma)… the only solution was to meet up online, through Google Hangouts.

The first meetings were incredibly boring and not concluding (sorry ITOUG guys if you read this :-)):

You cannot start something if you do not know what you want to achieve

Of course, everybody was agreeing that it would have been nice to become the new UKOUG in the southern Europe. But being realistic: no budget, no spare time, nothing at all, we needed a starting point. We though that the starting point was a website. But even something easy like a basic website forks a lot of additional questions:

  • Should it allow to publish content?
  • Should it have a bulletin board?
  • What about user subscription?
  • What should be the content? Articles, webinars?

We created something with the idea to publish our content, but after a while it was mostly an empty container.

It took me a while to learn a first big lesson:

Democracy does not work for small User Groups

The original founder of the group decided to quit the ITOUG because nothing concrete was happening. Everybody was putting ideas on the table but nothing was happening, really.

When my friend Björn Rost proposed me to candidate Milano for the OTN Tour, I jumped on the train: it was the best way to start something concrete. I somehow “forced” the OTN Tour candidature to happen, saying to my peers: “I will do it, if you support me, thanks; otherwise I will do it alone”.

And the response was great!

Do not wait for something to happen. If you want it, take the lead and make it.

This is the biggest lesson I have learned from my involvement with user groups. People sometimes cannot contribute because they do not what to do, how to do it, or simply they do not have time because there are a gazillion of things more important than the user groups: family, work, health… even hobbies sometimes are more important 🙂

If you have an idea, it’s up to YOU to transform it in something concrete.


After the acceptance, we had to prepare the event. We proposed a few dates and Björn prepared the calendar taking into account the other user groups.

Organizing an event is not easy but not so complex either

  • Set a reasonable target number of participants
  • Fix a date when most contributors are available
  • Get offers from different venues (pricing for the venue and the catering)
  • Get an idea of the budget
  • Ask different companies to sponsor the event
  • Eventually ask Oracle 🙂
  • Once the sponsors are committed to pay, block the venue
  • Prepare and publish the Call for Paper
  • Eventually start some advertising
  • Select the papers
  • Prepare the agenda
  • Ask the speakers for confirmation about the proposed date/time
  • Prepare the event registration form
  • Publish the agenda
  • Broadcast it everywhere 🙂 (Social media, contacts, website, Oracle through their channels)
  • Interact with the hotel and the sponsor to have the proper setup, billing addresses, invoices, etc.
  • Host the event
  • Relax

Finding the sponsors is the most difficult part

It has been easy for the first two events to find a sponsor (just a database stream, event held in Milano), but it was not the same for the last one.

Our aim was to do a double event (Milano + Roma) with two streams in each location (DB + BI & Analytics). In Roma we have been unable to find a sponsor (if you read this AND your company may be interested in sponsoring such event in Roma, please contact me :-)), we decided then to continue with the event in Milano.

Finding the speakers is easier than you can imagine

Unless you want non-english sessions held by native speakers, there is a huge community of speakers willing to share their knowledge. For Oracle, the most obvious source of speakers is the ACE Program, and twitter is probably the best channel for finding them.

Now it has been the third time that we organized an event, and every time we have been surprised by the good attendance and feedback.

A few images from the last event

DBygdizXkAE-EiZ DBynRIlXcAAqbtB DBy_0e8XUAECvW5

Which Oracle Databases use most CPU on my server?


  • You have many (hundreds) of instances and more than a couple of servers
  • One of your servers have high CPU Load
  • You have Enterprise Manager 12c but the Database Load does not filter by server
  • You want to have an historical representation of the user CPU utilization, per instance

Getting the data from the EM Repository

With the following query, connected to the SYSMAN schema of your EM repository, you can get the hourly max() and/or avg() of user CPU by instance and time.

Suppose you select just the max value: the result will be similar to this:


Putting it into excel

There are one million ways to do something more reusable than excel (like rrdtool scripts, gnuplot, R, name it), but Excel is just right for most people out there (including me when I feel lazy).

  • Configure an Oracle Client and add the ODBC data source to the EM repository:


  • Open Excel, go to “Data” – “Connections” and add a new connection:
    • Search…
    • New Source
    • DSN ODBC
  • Select your new ODBC data source, user, password
  • Uncheck “Connection to a specific table”
  • Give a name and click Finish
  • On the DSN -> Properties -> Definition, enter the SQL text I have provided previously


The result should be something similar: ( but much longer :-))

first_step_excelPivoting the results

Create e new sheet and name it “pivot”, Click on “Create Pivot Table”, select your data and your dimensions:

pivotThe result:

pivotedCreating the Graph

Now that the data is correctly formatted, it’s easyy to add a graph:

just select the entire pivot table and create a new stacked area graph.

The result will be similar to this:


With such graph, it is easy to spot which databases consumed most CPU on the system in a defined period, and to track the progress if you start a “performance campaign”.

For example, you can see that the “green” and “red” databases were consuming constantly some CPU up to 17.05.2017 and then some magic solved the CPU problem for those instances.

It is also quite convenient for checking the results of new instance caging settings…

The resulting CPU will not necessarily be 100%: the SYS CPU time is not included, as well as the user CPU of all the other processes that are either not DB or not monitored with Enterprise Manager.



Another problem with “KSV master wait” and “ASM file metadata operation”

My customer today tried to do a duplicate on a cluster. When preparing the auxiliary instance, she noticed that the startup nomount was hanging forever: Nothing in the alert, nothing in the trace files.

Because the database and the spfile were stored inside ASM, I’ve been quite suspicious…

The ASM trace files had the following entries:

The ASM instance had the following sessions waiting:


Around 12:38:56, another colleague in the office added a disk to one of the disk groups, through Enterprise Manager 12c!

But there were no rebalance operations:

It’s not the first time that I hit this type of problems. Sadly, sometimes it requires a full restart of the cluster or of ASM (because of different bugs).

This time, however, I have tried to kill only the foreground sessions waiting on “ASM file metadata operation”, starting with the one coming from the OMS.

Surprisingly, after killing that session, everything was fine again:

I never add disks via OMS (I’m a sqlplus guy ;-)) , I wonder what went wrong with it 🙂


RMAN Catalog Housekeeping: how to purge the old incarnations

First, let me apologize because every post in my blog starts with a disclaimer… but sometimes it is really necessary. 😉

Disclaimer: this blog post contains PL/SQL code that deletes incarnations from your RMAN recovery catalog. Please DON’T use it unless you deeply understand what you are doing, as it can compromise your backup and recovery strategy.

Small introduction

You may have a central RMAN catalog that stores all the backup metadata for your databases. If it is the case, you will have a database entry for each of your databases and a new incarnation entry for each duplicate, incomplete recovery or  flashback (or whatever).

You should also have a delete strategy that deletes the obsolete backups from either your DISK or SBT_TAPE media. If you have old incarnations, however, after some time you will notice that their information never goes away from your catalog, and you may end up soon or later to do some housekeeping. But there is nothing more tedious than checking and deleting the incarnations one by one, especially if you have average big numbers like this catalog:

Where db, dbinc, bdf and brl contain reslectively the registered databases, incarnations, datafile backups and archivelog backups.

Different incarnations?

Consider the following query:

You can run it safely: it returns the list of incarnations hierarchically connected to their parent, by database name, key and level.

Then you have several types of behaviors:

  • Normal databases (created once, never restored or flashed back) will have just one or two incarnations (it depends on how they are created):

They are usually the ones that you may want to keep in your catalog, unless the database no longer exist: in this case perhaps you omitted the deletion from the catalog when you have dropped your database?

  • Flashed back databases (flashed back multiple times) will have as many incarnations as the number of flashbacks, but all connected with the incarnation prior to the flashback:

Here, despite you have several incarnations, they all belong to the same database (same DB_KEY and DBID), then you must also keep it inside the recovery catalog.

  • Non-production databases that are frequently refreshed from the production database (via duplicate) will have several incarnations with different DBIDs and DB_KEY:

This is usually the most frequent case: here you want to delete the old incarnations, but only as far as there are no backups attached to them that are still in the recovery window.

  • You may also have orphaned incarnations:

In this case, again, it depends whether the DBID and DB_KEY are the same as the current incarnation or not.

What do you need to delete?


  • Incarnations of databases that no longer exist
  • Incarnations of existing databases where the database has a more recent current incarnation, only if there are no backups still in the retention window

How to do it?

In order to be sure 100% that you can delete an incarnation, you have to verify that there are no recent backups (for instance, no backups more rercent than the current recovery window for that database). If the database does not have a specified recovery window but rather a default “CONFIGURE RETENTION POLICY TO REDUNDANCY 1; # default”, it is a bit more problematic… in this case let’s assume that we consider “old” an incarnation that does not backup since 1 year (365 days), ok?

Getting the last backup of each database

Sadly, there is not a single table where you can verify that. You have to collect the information from several tables. I think bdf, al, cdf, bs would suffice in most cases.

When you delete an incarnation you specify a db_key: you have to get the last backup for each db_key, with queries like this:

Putting together all the tables:

Getting the  recovery window

The configuration information for each database is stored inside the conf table, but the retention information is stored in a VARCHAR2, either ‘TO RECOVERY WINDOW OF % DAYS’ or ‘TO REDUNDANCY %’

You need to convert it to a number when the retention policy is recovery windows, otherwise you default it to 365 days wher the redundancy is used. You can add a column and a join to the query:

and eventually, either display if it the incarnation is no more used or filter by usage:

Delete the incarnations!

You can delete the incarnations with this procedure:

This procedure will raise an exception (-20001, ‘Database not found’) when a database does not exist anymore (either already deleted by this procedure or by another session), so you need to handle it.

Putting all together:

I have used this procedure today for the first time and it worked like a charm.

However, if you have any adjustment or suggestion, don’t hesitate to comment it 🙂


Souvenirs from 2016

The 2016 is ending, at least from the Oracle Community point of view. It has been tiring and exciting at the same time, so I would like to put some good memories together.

This post is mostly for me, sorry 🙂

February: Another nice Tech Event

Trivadis Tech Event is a great conference, sadly not open for everyone, but still a great one… Got two (or three?) talks there.


March: a good beer in good company

Nearby the CERN, in Geneva, with a few good friends and big technologists:-)

cc9ymgkweaa1uxjMarch again: That ACE Director tweet

May: The DOAG Datenbank 2016

One speech there… the first of many about “upgrading 300 databases  in 300 days”. It was my first time speaking in Germany. 🙂

cilmlkzuoae99csMay again: The Italian leg of the OTN EMEA Tour

The OTN tour has been a great starter for the activities of the Italian Oracle User Group (of which I am one of the founders). It was great to discover that the interest for Oracle Database in Italy is still high (we got almost 60 people: that is huge for a first event, IMO).

We had Mark Rittman (before he became famous :-D), Christian Antognini, Frits Hoogland and Mike Dietrich!

cipw4ewwgauukzeSeptember: the ACED briefing, Oracle Open World and three spare days at Yosemite

It was my first time at the ACED briefind (key word: #cloud 😉 ) and also the first at Oracle HQ. It’s like going to Disney World, but the attractions are a little more scary 😀

cshime8uiaa4ygycsfozwrueaaq0g8The Yosemite was also incredible. In a single day of trekking, I scored 42k steps, 31km, +1200 vertical meters…

dsc09325dsc09426dsc09464dsc09490dsc09578dsc09594October: the great OTN Nordic Tour

That was fun, but incredibly tiring. 4 days in a row, 4 countries, 4 fligths, 4 different currencies, 4 ACE Directors and now 4 friends 🙂

I did not know very well Joel and Martin  and I did not know John at all. They are great people and I enjoyed a lot the time spent with them (and the beers :-D).





Stockholm was the last leg, I did it with John only. There I spent the rest of the week-end (the event was on Friday). I love Stockholm so much! Perhaps my favorite city (for sure in the top 5). I have also got a good whisky as speaker gift 😀

dsc_2343dsc_2402dsc09836dsc09920November: thesecond Italian Oracle User Group Event

We had again 60 people. In november I have also been speaker along with Christian Antognini, Mauro Pagano, Francesco Renne and Francesco Tisiot.

cw-izwiwiaaxosbNovember again: the DOAG

Definitely the best conference in Europe 🙂 It was my second time there and first one as speaker.

cxym09ouqaawfgxcxv6db2xeaamwqlNovember again: the Swiss Data Forum

It has been a great single-day event in Lausanne, not database centric but DATA centric, about Data, IoT, Big Data, Data Science, Deep Learning… I had one speech there.


December: the UKOUG Tech 16

Two final speeches at UKOUG in Birmingham. It was fun again, but the last day I did fell sick 🙁 (and some how I am still recovering).

czg9lvdusaahcikPlans for the 2017

I have got accepted for the IOUG Collaborate, but because of the many duties and all the recent travel, I have not confirmed my sessions (ouch, it is the first time that I do this, next time I will submit more carefully), so Open World will likely be my only US trip next year.

I look forward to submit for DOAG events again, speaking at SOUG (it’s already planned: 18th and 19th of May), and organizing at least two more events for the Italian Oracle User Group.

Happy New Year! 🙂

trivadis sessions at UKOUG Tech16

UKOUG Tech16 will start in less than a week. Trivadis will be there with many speakers, 10 sessions in total 🙂
If you are a delegate, come along and have a chat with us!

Super Sunday

Monday 05/12

Tuesday 06/12

Wednesday 07/12

See you there 🙂