[GENERAL] Misunderstanding deadlocks
I'm confused about how deadlock detection and breaking deadlocks works. Googling around it seems that the server should be detecting deadlocks and aborting one of the queries. But I'm getting occasional deadlocks that literally hang forever. I'm assuming they are deadlocks because they show up when running the queries I got from this url: https://wiki.postgresql.org/wiki/Lock_Monitoring I'm running postgres 9.3 on ubuntu, configuration is the default. Chris
[GENERAL] Postgres as key/value store
I'm looking for some feedback on the design I'm using for a basic key/value storage using postgres. Just some quick background. This design is for large scale games that can get up to 10K writes per second or more. The storage will be behind a distributed memory cache that is built on top of Akka, and has a write behind caching mechanism to cut down on the number of writes when you have many updates in a short time period of the same key, which is common for a lot of multiplayer type games. I have been using Couchbase, but this is an open source project, and Couchbase is basically a commercial product for all intents and purposes, which is problematic. I will still support Couchbase, but I don't want it have to tell people if you really want to scale, couchbase is the only option. The schema is that a key is a string, and the value is a string or binary. I am actually storing protocol buffer messages, but the library gives me the ability to serialize to native protobuf or to json. Json is useful at times especially for debugging. This is my current schema: CREATE TABLE entities ( id character varying(128) NOT NULL, value bytea, datatype smallint, CONSTRAINT entities_pkey PRIMARY KEY (id) ); CREATE OR REPLACE RULE entities_merge AS ON INSERT TO entities WHERE (EXISTS ( SELECT 1 FROM entities entities_1 WHERE entities_1.id::text = new.id::text)) DO INSTEAD UPDATE entities SET value = new.value, datatype = new.datatype WHERE entities.id::text = new.id::text; Additional functionality I want is to do basic fuzzy searches by key. Currently I'm using a left anchored LIKE query. This works well because keys are left prefixed with a scope, a delimiter, and then the actual key for the data. These fuzzxy searches would never be used in game logic, they would be admin only queries for doing things like obtaining a list of players. So they should be infrequent. The scope of the query ability will not expand in the future. I support multiple backends for the key/value storage so I'm working with the lowest common denominator. Plus I have a different approach for data that you need to do complex queries on (regular tables and an ORM). Chris
[GENERAL] Mirroring existing mysql setup
Where I work we use mysql for a fairly busy website, and I'd like to eventually start transitioning to postgres if possible. The largest obstacle is the lack of replication as a core feature. I'm well aware of the history behind why it's not in core, and I saw a post a while back saying it would be in 8.4. I'd like to say I think this is a very good idea, and I know a number of shops personally that did not go with postgres just for this reason. So anyways our general setup is that we have one master replicating to one slave. We use the slave for generating various leaderboard stats for our games. Most of these get generated every 4 hours. If we wanted to duplicate this on postgres I'm not sure which replication option would work best. Last time I looked at slony you had to edit configs for each table you wanted to replicate, and the whole setup was more complex then it needed to be. If it's still like that, I think we would lose more then we gain by moving to postgres. Once setup, the replication needs to be free of daily administration other then routine automated tasks. We add new tables/remove old ones almost on a daily basis. Now for one of the main things we don't like about mysql. You can't add indexes without locking the whole table, which means you can't go back and add indexes later on a production app without shutting down the whole system. The effect his has had is that when we add new features to our games that would normally require an additional column, we have to add a new table since we can't add an index to the old table. When you add indexes in postgres, how much of a performance hit will the database be taking while adding the index? I haven't worked on a postgres installation that's as busy as our mysql installation is. We get roughly 3-4 million page views per day, with each page view probably averaging 4-6 db queries. Probably 20% of these are cached. In addition we have our slave which does far fewer, but more complicated queries. Quite a few of our tables will gain thousands of rows per day, some tens of thousands. Some of our busiest tables have tens of millions of rows. We could start to archive some of these. Chris -- Sent via pgsql-general mailing list (pgsql-general@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
Re: [GENERAL] Reversing transactions on a large scale
On Thu, Nov 20, 2008 at 4:06 PM, Scott Marlowe [EMAIL PROTECTED] wrote: On Thu, Nov 20, 2008 at 4:36 PM, snacktime [EMAIL PROTECTED] wrote: Right now we are running mysql as that is what was there when I entered the scene. We might switch to postgres, but I'm not sure if postgres makes this any easier. We run a couple of popular games on social networking sites. These games have a simple economy,and we need to be able to time warp the economy back in time, which means reverting a whole lot of transactions and inventories. Our games generate around 1 million user transactions per hour, which results in inserts/updates on 4 times that many rows. Using PIT recovery would be a very reliable way to accomplish this, but I'm wondering how long it would take. If it takes a full day to roll back an hour of game time, then I need to find another solution. PITR is pretty fast, since it sequentially applies changes to the database as fast as it can. Your hardware has a lot to do with this though. Applying changes to a machine with plenty of memory, fast CPUs, and a big rockin RAID-10 array will of course be much faster than doing the same thing on a laptop. If you make base sets every night at midnight with snapshots, then it shouldn't take too long. Is this gonna be a regular thing, or is this more of an occasional occurance when things in the game go horribly wrong? It's primarily for when a bug screws up the economy, or if someone finds a way to hack the economy. Unfortunately these things happen now and then. Plus, these games are relatively short lived. We might get a million users the first month, but a year later the game is dead. So a generic solution using something like PITR would be good. It's not worth it to do it in code with the game having such a short lifespan. Chris -- Sent via pgsql-general mailing list (pgsql-general@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
[GENERAL] Reversing transactions on a large scale
Right now we are running mysql as that is what was there when I entered the scene. We might switch to postgres, but I'm not sure if postgres makes this any easier. We run a couple of popular games on social networking sites. These games have a simple economy,and we need to be able to time warp the economy back in time, which means reverting a whole lot of transactions and inventories. Our games generate around 1 million user transactions per hour, which results in inserts/updates on 4 times that many rows. Using PIT recovery would be a very reliable way to accomplish this, but I'm wondering how long it would take. If it takes a full day to roll back an hour of game time, then I need to find another solution. Chris -- Sent via pgsql-general mailing list (pgsql-general@postgresql.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
[GENERAL] Conditional updateable view
I can't seem to find an example of how to add restrictions to the where clause of an updateable view created via the rule system. For example I don't want the update to complete if a where clause is missing entirely, and in some cases I want to only allow the update if the where clause specifies a particular column. Is there a way to do this? Chris ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
[GENERAL] looking for some real world performance numbers
I'm working through the architecture design for a new product. We have a small group working on this. It's a web app that will be using ruby on rails. The challenge I'm running into is that the latest conventional wisdom seems to be that since obviously databases don't scale on the web, you should just not use them at all. I have a group of otherwise very bright people trying to convince me that a rdbms is not a good place to store relational data because eventually it won't scale. And of course we don't even have version 1 of our product out of the door. I'll admit we do have a very good chance of actually getting tons of traffic, but my position is to use a rdbms for relational data, and then if and when it won't scale any more, deal with it then. So what would really help me is some real world numbers on how postgresql is doing in the wild under pressure. If anyone cares to throw some out I would really appreciate it. Chris ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
[GENERAL] Modeling bill/ship addresses
I've been going back and forth on the best way to model this. A user can have one to many bill and ship addresses. An order can have one bill address and one to many ship addresses Let's assume I have a single address table, with an address_type column that is a foreign key to the address_types table. Now to create the relationships between addresses and users/orders. I could create a join table for holding the addresses that belong to orders. For example table order_addresses that has order_id and address_id columns that are foreign keys on addresses and orders. But what about just having two foreign keys in addresses? order_id and user_id? Or is there a rule against having a null foreign key? Also, is there a good database independent way to make the address immutable once it's created? I don't mind doing it at the application level actually, as I'm using a MVC framework that makes it easy to define that logic once in the model instead of spread out all over the codebase. Chris ---(end of broadcast)--- TIP 6: explain analyze is your friend
[GENERAL] server vendor recommendation
Any recommendations for vendors that can build custom servers? Specifically opteron based with scsi raid. Chris ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
[GENERAL] schema design question
Say you have 8 different data models that are related enough to share roughly 70% of the same fields, but the shared fields are not always the same. And also within any given model, some fields can be empty. The business logic is that data is pulled from all the data models and put into a common format that has all the combined fields, and sent over the wire as a transaction. The models are different credit card transaction types. Most of them share a lot of common fields, but each one is different by about 3-6 fields, and each one has different requirements on which fields can be empty and what their default values are. In the past I've basically used one table and modeled it after the message format that is sent to the processor. I'm thinking about creating a table for each transaction type though to cut down on the number of empty fields for each transaction type, and also to limit the number of options that is allowed in each field. For example across the transaction types credit, capture, authorize, force, charge, and voice authorize the authorization_source_code field might have 8 possible values, but in any one transaction type it probably only has 3 at the very most. Would most of you create a separate tabel for each transaction type? I'm also curious how others would handle the batch closings. In the past I have created separate tables for open transactions and transactions that have been captured/settled. When a transaction is captured it's moved to a different table instead of just having a column to mark it as captured. Normally I would select all the transactions to capture, insert them into the captured table, delete them from the open transactions table, process the batch, and if the batch goes through commit everything. That narrows down the number of things that can go wrong after you have submitted the batch. The alternative would be to just have a column to mark transactions as capture and leave them all in one table. I've always been paranoid about doing that because it leaves open the possibility of capturing thousands of transactions twice if you have a bug, as opposed to a few hundred at most. I spent quite a few years working at payment gateways and am now creating an open source platform that does the same thing that your normal payment gateway does. It's been a while since I've had the chance to look at this problem in a fresh light. Most of the processing code at payment gateways is left pretty much untouched once it's working, it's not something you go in and refactor every few months even if it's not perfect. Would appreciate any feedback. Chris ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
Re: [GENERAL] [Slony1-general] Using slony with many schema's
First, thanks for all the feedback. After spending some more time evaluating what we would gain by using slony I'm not sure it's worth it. However I thought I would get some more feedback before finalizing that decision. The primary reason for looking at replication was to move cpu intensive SELECT queries to a slave. However, by moving away from schema's the report queries for all clients on the server become more cpu intensive instead of just the clients with large data sets. The average distribution is that 95% of our clients have less then 5000 rows in any table, and the other 5% can have hundreds of thousands. So by putting all the data into one schema, every report query now gets run against a million or more rows instead of just a few hundred or thousand. So all clients will see a drop in query performance instead of just the clients with large amounts of data. Chris ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org/
Re: [GENERAL] [Slony1-general] Using slony with many schema's
Sorry wrong list, this was meant for the slony list... Chris ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [GENERAL] more anti-postgresql FUD
1. create table test (id int4, aaa int4, primary key (id)); 2. insert into test values (0,1); 3. Execute update test set aaa=1 where id=0; in an endless loop I just did the test on PostgreSQL 7.4.12 and MySQL 5.0.22 (MyISAM, sorry had no configured InnoDB). Ubuntu 6.0.6, AMD64, 2GB, default database settings. MySQL performs very well, approximately 15000-2 updates per second with no degradation of performance. PostgreSQL does approximately 1600 records per second for the first 1, then 200rps for the first 100k records, and then slower and slower downgrading to 10-20 rps(!!!) when reaching 300k. Something is wrong with your test code. If I had to guess I would say you did all the updates in a single transaction without committing them, in which case yes it will slow down until you commit. ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [GENERAL] using schema's for data separation
On 9/29/06, Just Someone [EMAIL PROTECTED] wrote: I am using a similar solution, and I tested it with a test containing 20K+ different schemas. Postgres didn't show slowness at all even after the 20K (over 2 million total tables) were created. So I have feeling it can grow even more. That's good to know we haven't really tested it against that many schema's, other then actually creating them to make sure there wasn't some sort of hard limit or bug no one had run into before. Performance with schema's is actually one thing I do like. A query for any one user is only hitting the data in the one schema, so users with large data sets don't impact the query performance of users with smaller data sets. ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
[GENERAL] using schema's for data separation
I'm re evaluating a few design choices I made a while back, and one that keeps coming to the forefront is data separation. We store sensitive information for clients. A database for each client isn't really workable, or at least I've never though of a way to make it workable, as we have several thousand clients and the databases all have to be accessed through a limited number of web applications where performance is important and things like persistant connections are a must. I've always been paranoid about a programmer error in an application resulting in data from multiple clients getting mixed together. Right now we create a schema for each client, with each schema having the same tables. The connections to the database are from an unprivileged user, and everything goes through functions that run at the necessary privileges. We us set_search_path to public,user. User data is in schema user and the functions are in the public schema. Every table has a client_id column. This has worked well so far but it's a real pain to manage and as we ramp up I'm not sure it's going to scale that well. So anyways my questions is this. Am I being too paranoid about putting all the data into one set of tables in a common schema? For thousands of clients what would you do? Chris ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
[GENERAL] Kill specific connection
What's a safe way to kill a specific connection to the database? I'm testing some code that reconnects if a connection has timed out or gone bad and I need to simulate a connection that has gone away. Chris ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
[GENERAL] implementing a read lock
I have an application that processes credit card transactions,and contains a table called authorizations. The authorizations table contains information returned by the bank necessary to capture the transaction. Nothing should block the application from inserting new rows into the authorizations table. When the authorizations are captured, one or more rows will be fetched, captured, and if successful the rows will be deleted. No updates are done on the table, only inserts or deletes. Naturally I want to prevent different instances of the same application from trying to select the same rows to capture, resulting in duplicate charges. I can't lock the whole table because new authorizations are constantly coming in. Is creating a separate table that I use just as a lock table the best approach? ---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster
[GENERAL] alternative to using a sequence
I have an application that processes financial transactions. Each of these transactions needs to be sent with a sequence number. It starts at 1 and resets to 1 once it hits 8000. I'm trying to think of the most elegant solution without having to create a sequence for each user (there are hundreds). There is a table that holds the configuration parameters for each merchant, so a field in that table to hold the sequence number would be ideal. In the past I've used sequences as well as just a field which I query then update. Any other ideas? ---(end of broadcast)--- TIP 6: explain analyze is your friend
Re: [GENERAL] User privileges in web database applications
Both connection pooling and using the superuser with SET SESSION AUTHORIZATION both have their uses. You might have an application that processes some type of transaction and inserts data into a users schema or table, but where there are no user credentials available. Then you might have a web interface for users to access that data where user credentials are available. We have this type of setup and we use a superuser with SET SESSION AUTHORIZATION for the incoming transactions.But we also have extensive security requirements that demand we do things most people don't do. Full security/code audits every quarter, peer review and full testing for any new code, hardware encryption for sensitive data and keys stored on tokens, client certificate authentication for all web access, restrictive firewall, etc.. Bottom line is that I'm paranoid about using SET SESSION AUTHORIZATION, but it does have it's uses and can be used safely. ---(end of broadcast)--- TIP 5: don't forget to increase your free space map settings
Re: [GENERAL] minimizing downtime when upgrading
On 6/16/06, Richard Huxton dev@archonet.com wrote: The other option would be to run replication, e.g. slony to migrate from one version to another. I've done it and it works fine, but it will mean slony adding its own tables to each database. I'd still do it one merchant at a time, but that should reduce your downtime to seconds. I'll have to take another look at slony, it's been a while. Our database structure is a bit non standard. Being a payment gateway, we are required to have a separation of data between merchants, which means not mixing data from different merchants in the same table. So what we do is every user has their own schema, with their own set of tables. Yes I know that's not considered the best practice design wise, but separate databases would have caused even more issues, and as it turns out there are some advantages to the separate schema approach that we never thought of. Last time I looked at slony you have to configure it for each individual table you want replicated. We have around 50,000 tables, and more are added on a daily basis. ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
[GENERAL] minimizing downtime when upgrading
Anyone have any tips for minimizing downtime when upgrading? So far we have done upgrades during scheduled downtimes. Now we are getting to the point where the time required for a standard dump/restore is just too long. What have others done when downtime is critical? The only solution we have been able to come up with is to migrate the data on a per user basis to a new database server. Each user is a merchant, and the data in the database is order data. Migrating one merchant at a time will keep the downtime per merchant limited to just the time it takes to migrate the data for that merchant, which is acceptable. Any other ideas? ---(end of broadcast)--- TIP 3: Have you checked our extensive FAQ? http://www.postgresql.org/docs/faq
Re: [GENERAL] Dynamic web sites with PostgreSQL
On 11/9/05, Nicolay A Vasiliev [EMAIL PROTECTED] wrote: Hello there!I'd like to ask the PostgreSQL community for the conseptual thing. Wedevelop our web sites using MySQL. We like this for its high speed andfulltext search feature. But nowadays our projects are growing fast and we afraid our MySQL won't be able to perform large amount of complexcomplicated queries. So we get a question about altenative SQL server.In fact there are not too much from open source SQL servers, I think only 2 serious: PostgreSQL and MaxDB. May I sak you about words foradvocacy or accusation for each of these database servers? Run as fast as you can away from MaxDB. I've looked at the source in detail, and my money is that mysql ab will never really do anything with it. It's just a huge mess. The documentation is probably the worst I have ever seen, the command line utilities have so many options and are so poorly documented that it takes forever to figure out how to do the most basic stuff. Exporting and importing data is also a pain. I really can't say enough bad things about MaxDB. Maybe it was good a few years ago, but right now it's a huge step backwards from most anything else you could use, including mysql. Chris
Re: [GENERAL] Oracle buys Innobase
On 10/7/05, Jim C. Nasby [EMAIL PROTECTED] wrote: http://lnk.nu/prnewswire.com/4dv.pl--Jim C. Nasby, Sr. Engineering Consultant[EMAIL PROTECTED]Pervasive Software http://pervasive.comwork: 512-231-6117vcard: http://jim.nasby.net/pervasive.vcf cell: 512-569-9461---(end of broadcast)--- TIP 2: Don't 'kill -9' the postmaster First thing that comes to my mind is that Oracle is setting the stage to buy them out.
Re: [GENERAL] per user logging
I believe it is possible for a superuser to do something likeALTER USER victim SET log_min_messages = whatever, so that the log verbosity is different for different users.regards, tom lane I'll try that and see how it works.
[GENERAL] per user logging
I'm pretty sure the answer to this is no, but just in case I've missed something. Is there a way to configure the server so it only logs for specific users? Chris
[GENERAL] changing default lockfile location
I'm trying to run two database clusters on the same box. Both are bound to their own ip but use the same port. I can't see a way to change the location of the lockfile on a per cluster basis though. Is there one? Chris ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
Re: [GENERAL] changing default lockfile location
On 8/5/05, snacktime [EMAIL PROTECTED] wrote: I'm trying to run two database clusters on the same box. Both are bound to their own ip but use the same port. I can't see a way to change the location of the lockfile on a per cluster basis though. Is there one? Chris Never mind, I figured out the socket file is also the lockfile. Chris ---(end of broadcast)--- TIP 4: Have you searched our list archives? http://archives.postgresql.org
Re: [GENERAL] Transparent encryption in PostgreSQL?
On 7/13/05, Matt McNeil [EMAIL PROTECTED] wrote: Greetings, I need to securely store lots of sensitive contact information and notes in a freely available database (eg PostgreSQL or MySQL) that will be stored on a database server which I do not have direct access to. This database will be accessed by a PHP application that I am developing. However, I also need to be able to search/sort these data with the database functions (SELECT, ORDER BY, etc) so encrypting on the client side (web application) or using encryption of specific fields would not work. (For example, I need to encrypt contacts' names, but need to be able to search for results by name). (I realize I could load the entire table into memory with PHP and process/search/sort it there, but that's obviously not a very good solution). Ideally I would like to encrypt entire tables. I read something about the pgcrypto contrib module, but have't been able to discern if it can do ecryption in a transparent way (e.g. so that I can do regex searches on the data). My sense is that this is a difficult problem. However, I made the mistake of promising this functionality, so I'm scrambling to figure out some kind of solution. Any suggestions? Go back to your client and renegotiate? That would be my first option. Encryption isn't all it's cracked up to be, especially when most people think they can ignore application level security just by encrypting their data. One possible compromise might be to lower case the name, strip out any spaces or punctuation, split it into first and last name (if it's web based use separate input fields), and then hash those values and stick them in their own rows. You would also have the full name encrypted in it's own row. To search you lower case the search string, strip spaces, hash the value, and perform the query. Kind of a hack but it works. Chris ---(end of broadcast)--- TIP 1: if posting/reading through Usenet, please send an appropriate subscribe-nomail command to [EMAIL PROTECTED] so that your message can get through to the mailing list cleanly
[GENERAL] strange problem with pl/pgsql function caching of bad values
I have a very strange issue that I'm not sure how to debug. This is on postgresql 8.0.0rc5, Freebsd 5.4. Yes I know I should be upgrading this version and it's scheduled, but it can't happen for another week and for all I know this might still be an issue in current versions of postgresql. First the function in question: CREATE OR REPLACE FUNCTION cancel_subscription_bysubid(varchar) returns integer AS ' DECLARE in_s_oid varchar; in_active varchar; status integer; BEGIN in_s_oid := $1; in_active := 0; status := active from recurbilling_transactions where s_oid = in_s_oid; IF status = 0 THEN RETURN 0; ELSIF status = 1 THEN EXECUTE ''update recurbilling_transactions set active= '' ||in_active|| '' WHERE s_oid ='' || quote_literal(in_s_oid); RETURN 1; ELSE RETURN 2; END IF; END ' LANGUAGE 'plpgsql' SECURITY DEFINER; The 'active' column is an integer NOT NULL. s_oid is a varchar. Every few days the database gets into a state where this function starts returning a value of 2 even though the value of 'active' is 1 or 0. Even stranger is that not all sessions will do this. We used cached connections via the perl DBI, and once this starts happening some sessions return the bad value while others work correctly. Once the database is in this state testing the function via psql at the command line will always result in the function returning a value of 2, while some of the perl DBI connections will still be returning 0 or 1. I'm assuming that at some point all new sessions get hosed and it's the older sessions which still work. Restarting the database puts things back to normal and the function then works correctly again for a while. One other thing about our particular setup is that we use separate schema's for all user data and the functions go in the public schema. So before executing this function we issue something like 'set_path to username,public'. Chris ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
[GENERAL] upgrading from 8.0 rc5
Do I need to do a full dump/restore when migrating from 8.0 rc5 to the latest 8.0.3? Chris ---(end of broadcast)--- TIP 9: In versions below 8.0, the planner will ignore your desire to choose an index scan if your joining column's datatypes do not match
Re: [GENERAL] strange problem with pl/pgsql function caching of bad values
One other thing about our particular setup is that we use separate schema's for all user data and the functions go in the public schema. So before executing this function we issue something like 'set_path to username,public'. Mph. Are you expecting the function to work for more than one such path value over the life of a connection? Maybe you need to do the selection part with an EXECUTE not only the update. As-is, the first execution will latch down which copy of recurbilling_transactions will be used for the selection, regardless of later changes in search_path. Argh... I knew better. Yes i need to use EXECUTE because it is used across multiple paths. Chris ---(end of broadcast)--- TIP 3: Have you checked our extensive FAQ? http://www.postgresql.org/docs/faq
