On 18-10-2017 09:18, Ivan Sagalaev wrote:
Hello everyone,
An inaugural poster here, sorry if I misidentified a list for my
question.
I am planning to use PostgreSQL as a storage for application logs
(lines of text) with the following properties:
- Ingest logs at high rate: 3K lines per second minimum, but the more
the better as it would mean we could use one Postgres instance for
more than one app.
- Only store logs for a short while: days, may be weeks.
- Efficiently query logs by an arbitrary time period.
- A "live feed" output, akin to `tail -f` on a file.
For context, I only used Postgres for a bog standard read-heavy web
apps, so I'm completely out of expertise for such a case. Here are my
questions:
- Is it even possible/advisable to use an actual ACID RDBMS for such a
load? Or put another way, can Postgres be tuned to achieve the
required write throughput on some mid-level hardware on AWS? May be at
the expense of sacrificing transaction isolation or something…
- Is there an efficient kind of index that would allow me to do `where
'time' between ... ` on a constantly updated table?
- Is there such a thing as a "live cursor" in Postgres for doing the
`tail -f` like output, or I should just query it in a loop (and skip
records if the client can't keep up)?
Thanks in advance for all the answers!
Hello,
not much on the topic, I had the same problem and I solved it by using a
Redis server (memory is cheap and fast) to store the logs for an
hour / day depending on the load average and then drop them on a csv or
sql file and insert it into Postgresql database.
My Redis record is so structured that I have the ability to review the
current actions of each user like tail -f.
Hardware is not much, Redis server with a lot of memory and cheap server
for database to store logs and I now even try to make different approach
to
remove the database server, because I store every day as separate gziped
log file for backup.
Regards,
Hristo S
--
Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general
