Log of /couchdb/trunk/src/couchdb/couch_rep_writer.erl

---

Revision 882903 - (view) (annotate) - [select for diffs]
Modified Sat Nov 21 13:43:43 2009 UTC (2 days, 23 hours ago) by kocolosk
File length: 2706 byte(s)
Diff to previous 817403 (colored)

code improvements from tidier. Patch by Kostis Sagonas. COUCHDB-570

---

Revision 817403 - (view) (annotate) - [select for diffs]
Modified Mon Sep 21 20:29:22 2009 UTC (2 months ago) by kocolosk
File length: 2761 byte(s)
Diff to previous 815921 (colored)

attempt at gentler replication shutdown when one of the DBs is deleted

---

Revision 815921 - (view) (annotate) - [select for diffs]
Modified Wed Sep 16 18:37:54 2009 UTC (2 months, 1 week ago) by kocolosk
File length: 2584 byte(s)
Diff to previous 807208 (colored)

GC isn't happening for this process b/c heap is too small, so trigger it manually

---

Revision 807208 - (view) (annotate) - [select for diffs]
Modified Mon Aug 24 13:31:48 2009 UTC (3 months ago) by kocolosk
File length: 2549 byte(s)
Diff to previous 803303 (colored)

skip full_commit during replication when its safe

---

Revision 803303 - (view) (annotate) - [select for diffs]
Modified Tue Aug 11 21:14:53 2009 UTC (3 months, 1 week ago) by kocolosk
File length: 2462 byte(s)
Diff to previous 803220 (colored)

roll back streaming _bulk_docs b/c of a race condition

---

Revision 803220 - (view) (annotate) - [select for diffs]
Modified Tue Aug 11 17:25:13 2009 UTC (3 months, 1 week ago) by kocolosk
File length: 5465 byte(s)
Diff to previous 802987 (colored)

more work on _bulk_docs streaming during replication.

I think this effort is kind of a dead end, at least if we're serious about
that 4GB maximum document/attachment size.  Presently push replication will
have trouble with attachments larger than a few MB because of the
inactivity_timeout we've set in ibrowse.  We'll push the data, but then close
the connection after 30 seconds while the target is still decoding.

We should focus our efforts instead on a way to push attachments without Base64.
I don't know what to say about 4GB _documents_, I don't see how we could ever
really support those now even without replication.

---

Revision 802987 - (view) (annotate) - [select for diffs]
Modified Tue Aug 11 04:51:03 2009 UTC (3 months, 2 weeks ago) by kocolosk
File length: 3739 byte(s)
Diff to previous 802986 (colored)

whoops, process dictionary certainly won't work there

---

Revision 802986 - (view) (annotate) - [select for diffs]
Modified Tue Aug 11 03:47:25 2009 UTC (3 months, 2 weeks ago) by kocolosk
File length: 3554 byte(s)
Diff to previous 802983 (colored)

split into multiple requests so target doesn't buffer too much

---

Revision 802983 - (view) (annotate) - [select for diffs]
Modified Tue Aug 11 02:54:42 2009 UTC (3 months, 2 weeks ago) by kocolosk
File length: 3108 byte(s)
Diff to previous 802974 (colored)

max chunk size of 4G, not 64k

---

Revision 802974 - (view) (annotate) - [select for diffs]
Modified Tue Aug 11 01:51:06 2009 UTC (3 months, 2 weeks ago) by kocolosk
File length: 3108 byte(s)
Diff to previous 802888 (colored)

stream _bulk_docs JSON body during replication

---

Revision 802888 - (view) (annotate) - [select for diffs]
Added Mon Aug 10 18:37:43 2009 UTC (3 months, 2 weeks ago) by kocolosk
File length: 2625 byte(s)

new replicator using _changes feed for continuous replication