| From: | Tomas Vondra <tv(at)fuzzy(dot)cz> |
|---|---|
| To: | Chris Browne <cbbrowne(at)acm(dot)org> |
| Cc: | pgsql-general(at)postgresql(dot)org |
| Subject: | Re: INSERT performance deteriorates quickly during a large import |
| Date: | 2007-11-09 20:59:55 |
| Message-ID: | 4734CA4B.3040102@fuzzy.cz |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-general |
>> Try to one of these:
>>
>> a) don't use INSERT statements, use a COPY instead
>>
>> b) from time to time run ANALYZE on the "public" table (say 1000
>> inserts, then one analyze)
>>
>> c) create the table without constraints (primary / foreign keys in this
>> case), import all the data, and then create the constraints
>>
>> The (b) and (c) may be combined, i.e. import without constraints and
>> analyze from time to time. I'd probably try the (a) at first, anyway.
>>
>> Try to gather some more statistics - is the problem related to CPU or
>> I/O? Use 'dstat' for example - this might say give you a hint in case
>> the advices mentioned above don't help.
>
> I agree with all but b).
You're right - this combination (no constraints or indices and using
ANALYZE at the same time) won't help. The original list of advices was a
little bit different, but I've changed it and haven't checked the
following paragraphs ...
Tomas
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Ted Byers | 2007-11-09 21:18:17 | Re: Optimal time series sampling. |
| Previous Message | Nathan Wilhelmi | 2007-11-09 20:00:41 | Importance of CPU floating point performance... |