Maintaining an index on a large table: Is there any potential for it to stall my application? - Mailing list pgsql-general

From Peter Geoghegan
Subject Maintaining an index on a large table: Is there any potential for it to stall my application?
Date
Msg-id db471ace0806021126n20107cd3g9259cc05e9b1d591@mail.gmail.com
Whole thread Raw
Responses Re: Maintaining an index on a large table: Is there any potential for it to stall my application?  (Craig Ringer <craig@postnewspapers.com.au>)
List pgsql-general
Hello,

I'm writing a C++ application that stores data in a table that may
ultimately become very large (think tens of millions of rows). It has
an index on one row, in addition to the index created on/as part of
its primary key. My concern is that a call to the pl/pgSQL function
that INSERTs data into this table might eventually lock the UI for an
annoyingly long time, as control flow in my application waits for that
Pl/PgSQL function to return.

I suspect that maintaining a large index could eventually cause this
to happen, particularly as the hardware that this embedded application
will run on is typically modest. Can someone suggest a solution? Are
my concerns about this happening well founded? At the moment, the
application doesn't lock at all, but it only has a few thousand rows.
I'm aware that I could create a second thread to make the call to my
database API, libpqxx, but I have reservations due to the possible
implications for thread safety - pqxx lacks "a flexible mechanism for
thread synchronization", so this might cause headaches.

Thanks,
Peter Geoghegan

pgsql-general by date:

Previous
From: "Henry"
Date:
Subject: Re: dblink() cursor error/issue (TopMemoryContext)
Next
From: John Cieslewicz
Date:
Subject: Forcing Postgres to Execute a Specific Plan