log file truncation

[janl]

Helps find corrupted files on servers with lots of shards.

test script:

#!/bin/sh

COUCH=https://a:[email protected]:15984

DB=x-test-goop
CURL="curl -s"

# file does not exist
ls ~/Work/asf/couchdb/dev/lib/node1/data/shards/00000000-ffffffff/$DB*.couch

# reset db
$CURL -X DELETE $COUCH/$DB
$CURL -X PUT $COUCH/$DB?q=1

# add docs so we have something to truncate
$CURL -X PUT $COUCH/$DB/doc1 -d '{"foo":"baaaaaah1"}'
$CURL -X PUT $COUCH/$DB/doc2 -d '{"foo":"baaaaaah2"}'
$CURL -X PUT $COUCH/$DB/doc3 -d '{"foo":"baaaaaah3"}'

# show that we can read from the db
$CURL $COUCH/$DB/_changes?include_docs=true

# truncate
FILE_NAME=`ls ~/Work/asf/couchdb/dev/lib/node1/data/shards/00000000-ffffffff/$DB*.couch`
FILE_SIZE=`ls -l $FILE_NAME | awk '{print $5}'`
NEW_FILE_SIZE=`echo "$FILE_SIZE-5000" | bc`
truncate -s $NEW_FILE_SIZE $FILE_NAME
# rm $FILE_NAME
REAL_NEW_FILE_SIZE=`ls -l $FILE_NAME | awk '{print $5}'`

echo "$FILE_NAME"
echo "fs: $FILE_SIZE, nfs: $NEW_FILE_SIZE, rnfs: $REAL_NEW_FILE_SIZE"

# show error
$CURL $COUCH/$DB/_changes?include_docs=true

Response without the patch:

{"error":"error","reason":"{{{badmatch,eof},\n  [{couch_file,read_raw_iolist_int,3,[{file,\"src/couch_file.erl\"},{line,714}]},\n   {couch_file,handle_call,3,[{file,\"src/couch_file.erl\"},{line,493}]},\n   {gen_server,try_handle_call,4,[{file,\"gen_server.erl\"},{line,636}]},\n   {gen_server,handle_msg,6,[{file,\"gen_server.erl\"},{line,665}]},\n   {proc_lib,init_p_do_apply,3,[{file,\"proc_lib.erl\"},{line,247}]}]},\n {gen_server,call,[<0.784.0>,{pread_iolist,16830},infinity]},\n [{gen_server,call,3,[{file,\"gen_server.erl\"},{line,214}]},\n  {couch_file,pread_iolist,2,[{file,\"src/couch_file.erl\"},{line,177}]},\n  {couch_file,pread_binary,2,[{file,\"src/couch_file.erl\"},{line,172}]},\n  {couch_file,pread_term,2,[{file,\"src/couch_file.erl\"},{line,160}]},\n  {couch_btree,get_node,2,[{file,\"src/couch_btree.erl\"},{line,435}]},\n  {couch_btree,reduce_stream_node,11,\n               [{file,\"src/couch_btree.erl\"},{line,622}]},\n  {couch_btree,fold_reduce,4,[{file,\"src/couch_btree.erl\"},{line,81}]},\n  {couch_bt_engine,count_changes_since,2,\n                   [{file,\"src/couch_bt_engine.erl\"},{line,646}]}]}"}

Response with the patch:

{"error":"error","reason":"{{bad_return_value,\n     {file_truncate_error,eof,\n         \"/Users/jan/Work/asf/couchdb/dev/lib/node1/data/shards/00000000-ffffffff/x-test-goop.1618391080.couch\"}},\n {gen_server,call,[<0.6608.1>,{pread_iolist,16830},infinity]},\n [{gen_server,call,3,[{file,\"gen_server.erl\"},{line,214}]},\n  {couch_file,pread_iolist,2,[{file,\"src/couch_file.erl\"},{line,177}]},\n  {couch_file,pread_binary,2,[{file,\"src/couch_file.erl\"},{line,172}]},\n  {couch_file,pread_term,2,[{file,\"src/couch_file.erl\"},{line,160}]},\n  {couch_btree,get_node,2,[{file,\"src/couch_btree.erl\"},{line,435}]},\n  {couch_btree,reduce_stream_node,11,\n      [{file,\"src/couch_btree.erl\"},{line,622}]},\n  {couch_btree,fold_reduce,4,[{file,\"src/couch_btree.erl\"},{line,81}]},\n  {couch_bt_engine,count_changes_since,2,\n      [{file,\"src/couch_bt_engine.erl\"},{line,646}]}]}"}

[bessbd]

The error message improvement seems awesome, but we still shouldn't get a stack trace on HTTP, right?

[bessbd]

Should we open a follow-up issue / PR?

[janl]

@rnewson is this more in line with what you were looking for here?

[rnewson]

Yes, sorry for delay. +1.