CouchDB 2.3 changed format of fields update_seq and purge_seq.
This causes a problem when Fabric maps the CouchDB
JSON resposnes to structs.

Since Fabric doesn't use these fields anyways, they can be removed
from the Fabric CouchDB structs. This will make Fabric
compatible with CouchDB 2.3.

Change-Id: I89f789ef79ea485aa74b99ef047af417a8be044e
Signed-off-by: David Enyeart <enyeart@us.ibm.com>

Added health check for CouchDB

Change-Id: If2d32b2197162b5f04997ab71ba2d292701cd0b9
Signed-off-by: Saad Karim <skarim@us.ibm.com>
Signed-off-by: Matthew Sykes <sykesmat@us.ibm.com>
(cherry picked from commit 8768567a9c626b2feb533f3c4909d2243d44aac9)

Using, attachments=true, fetch all attachments in a CouchDB range
queries instead of first obtaining metadata and fetching each attachment
in a separate call.

Change-Id: Ibe51e621d9c49116ebd9fbd866f939eecff326a4
Signed-off-by: ChanderG <mail@chandergovind.org>

ReadDoc function misses out setting the Length field in the
attachments in CouchDoc struct

Change-Id: Ife6d4206cf428031ab00392dd90a34e4404325d9
Signed-off-by: ChanderG <mail@chandergovind.org>

Change-Id: I0a08b7f47c1f6c2ecbcca204d3c5bcb04c6858d8
Signed-off-by: Matthew Sykes <sykesmat@us.ibm.com>

Metric added that will record the time a request to couch db takes to
complete.

Change-Id: Ie2068bb7b97a7103306093a2a6618ed0caed6b4b
Signed-off-by: Saad Karim <skarim@us.ibm.com>
Signed-off-by: Matthew Sykes <sykesmat@us.ibm.com>

The signature of handleRequest update to allow for passing in database
name, query parameters, and path elements.

Construction of the request URL now happens inside of handleRequest,
rather than relying on caller to construct the URL.

Change-Id: I2ac8ea273ebc71e29947261601c136db00c9bfd7
Signed-off-by: Saad Karim <skarim@us.ibm.com>
Signed-off-by: Matthew Sykes <sykesmat@us.ibm.com>

Improve ledger debug based on serviceability findings.

- Remove unnecessary debug.
- Add info logging for state and history rebuild.
- Add couchdb database name to couchdb debug.

Change-Id: Ia39b63d55529ed9ee40325b3a03a8bf78cb667ed
Signed-off-by: David Enyeart <enyeart@us.ibm.com>

Currently the warning is only seen in peer logs if couchdb debug is enabled.

Peer should log a warning without having to enable couchdb debug.

Log warning to peer when "no matching index found" warning comes from
couchdb queries.

Change-Id: I46a5fa663c502bbdf74261a4925ce54633384d1f
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Currently, CouchDB queries are limited to configurable max number of
results/docs, based on peer queryLimit config option (default 10000).
The shim/peer supports pagination but CouchDB query iterators do not.

This change will allow additional controls to prevent a large accidental or
malicious query from causing performance issues with CouchDB.

Changes to core.yaml:

Changed queryLimit to 1000.  This is part of repurposing this parameter.
QueryLimit is now to be used as a per query limit to CouchDB instead of the
limit for the query from the shim.

Added totalLimit parameter.  This parameter will be used to cap the maximum
number of records returned by the query from the shim.

Changes to couchdb:

Added changes to range query to find the next start key (if exists) for the
specified range.  The next start key is returned with the query results.

Added changes to rich query to find the next bookmark for the query.  The
bookmark is returned with the query ...

Replace go-logging with a zap based implementation. The implementation
is mostly compatible with the legacy format but differs in a two key
ways:

1. CRITICAL and NOTICE log levels no longer exist. Calls to Critical get
   mapped to error and NOTICE gets mapped to INFO. The methods
   associated with these levels will be removed in a future CR.
2. The log level constants are now sourced from the zap/zapcore packages.
   These values are incompatible wit the go-logging constants. Please be
   sure to use the appropriate constant when necessary.

The existing go-logging package is still used by the chaincode shim to
reduce the risk of negatively impacting legacy chaincode.

Change-Id: Iaf5fac807244883a8285892ccd350c5256959782
Signed-off-by: Matthew Sykes <sykesmat@us.ibm.com>

Change-Id: Idcee6cc17854db52c3a5b38f162171742c47d2d6
Signed-off-by: Firas Qutishat <firas.qutishat@securekey.com>

The CouchDB implementation currently does not allow + chars in the
database name.

Add support in the CouchDB implementation layer for the + char
be included as part of the database name.

Add unit tests for databases containing a + char.

Change-Id: I59001d1aed19066d65f6e230f41958cd2a8cbd73
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Part 2 of ledger servicebility improvement:
Update core/ledger to use error handling framework

Change-Id: I5c97efdeb071529c4da511491656bed8e947930a
Signed-off-by: Wenjian Qiao <wenjianq@gmail.com>

An error returned by CouchDB with a 500 error should
return the CouchDB error instead of the golang return code.

Change-Id: I7ecd085fbc2f87baae20db7c216419a373cdf47b
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Add changes to add database security to each CouchDB database.   This will
prevent any unauthorized browsing of state data.

Add couchdb methods to expose the CouchDB _security API for PUT and GET to
respectively create and retrieve the database security permissions.

Add the following structure to couchdb to represent the couchdb security
structure:

type DatabaseSecurity struct {
	Admins struct {
		Names []string `json:"names"`
		Roles []string `json:"roles"`
	} `json:"admins"`
	Members struct {
		Names []string `json:"names"`
		Roles []string `json:"roles"`
	} `json:"members"`
}

Add changes to couchdb CreateDatabaseIfNotExist() and apply a security
permission definition based on the configured CouchDB username.  Add the
username to the Admins and Members section.

Change-Id: I8dadd0e1575babef2143c7fb332b863aa0c6be5b
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Change-Id: I6684440cacb139ec70c54bd6d285d769f60d1b8d
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>
Signed-off-by: Ry Jones <ry@walledcity.org>

Problem Statement:

If the provided index definitions have more than one (lets say N)
index with same DesignDoc and indexName but with potentially different indexes,
the log will contain N messages like `created CouchDB index in the channel...`
but in reality only 1 index will be created in CouchDB. Fabric goes through
each file in alphabetic order, creating and (quietly) replacing the index
created in last round with the same DDoc and indexName, so only the last index
remains in CouchDB.

Solution:

Part of the confusion can be prevented by updating the message when an index
is created or updated in CouchDB.  The create index command responds with
metadata that can be captured and provide a more detailed message and can
determine if the index has been created or updated.  This can be accomplished
with the existing http request.

Change-Id: I84674ceed54dadba9470d768a028a6acb3b237f6
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

The CouchDB retry currently assumes the first pass through is
actually retry number 1.

It also does not currently allow a setting of zero retries.
This can be confusing to the end user and will prevent some potential
use cases that may need to wrap the request handler with zero retry logic.

Change-Id: I74a5628f19fd6d73a24091bad66f955046e28190
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

CouchDB index warming is executed per block, and as a result it
may become an unexpected resource hog in some scenarios, such
as when a high rate of updates going on with very small Batchsize.

A new config option warmIndexesAfterNIntervals has been added to
prevent index warming from happening on every block.  Default is
configured as 1.

Change-Id: If7198a4140e27d312ee7caa2f667bf2c80df3657
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Upon code review in statecouchdb.go and couchdb.go, it appears
there are occurrences of the following error producing functions
without errors captured and handled:

json.Marshal()
json.Unmarhsal()
decoder.Decode()
   e.g. json.NewDecoder(resp.Body).Decode(&dbResponse)
   e.g. decoder.Decode(&genericMap)

Change-Id: Id632d636d7e2de817abb60961f908f7ad5222bec
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

In rocket.chat there was a post about an error:
when i am running join peer request some times it's giving
Error:file_exists, Status Code:412, Reason:The database could
not be created, the file already exists

I believe this can happen if peer is retrying a create database
request against CouchDB.  A prior attempt to create database came
back as a timeout failure to peer, but in the meantime the database
got created in CouchDB.  The next retry would get the error above.
This scenario should be handled in peer code.

Change-Id: I01ecd3a5fde8a9042f6a18181376dec267d0cc15
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Chaincode data is saved in CouchDB under a JSON "data" envelope.

For queries, peer automatically re-writes chaincode query to account for
data envelope.

For indexes, chaincode author must specify the data envelop when referencing
fields.

Need to make the experience consistent, by removing the "data" envelope.
This will fix the usability issue, and also simplify the code, eliminating
other possible defects.

This change is only possible in 1.1, since the upgrade from 1.0 to 1.1 will
rebuild CouchDB state database using the enhanced 1.1 CouchDB data structure.

Change-Id: I2f7345fe4f4aaa17324c26f1e32bec2b3a5f67db
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Per review comments in https://gerrit.hyperledger.org/r/#/c/15057/

 ,
the index warming in the commit path will impact performance, it needs
to be moved out of commit path in a separate goroutine or daemon.

This change executes the WarmIndexAllIndexes() as a background process.

Change-Id: I706cae02d1e07551849ae301381f2b3ff218ade8
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Chaincode event handling was intended to pick up LSCC
chaincode updates, but was also picking up LSCC
chaincode collection updates.

This change ignores chaincode collection updates.

Additionally, related chaincode event debug is improved.

Note that two copies of Collecion-specific constants
were found.  The unused duplicate is removed from lscc.

Change-Id: I8d7c4320d757a7d3327bd2bb22bba0b653febc23
Signed-off-by: David Enyeart <enyeart@us.ibm.com>

CouchDB does not automatically update indexes when records are
inserted or updated.  The update is performed for the index
during the query operation.   This can cause delays during
the query.   Subsequent queries then take advantage of the updated
index.

This improvement will add an "index warming" step to refresh indexes
following an insert or update.

The addition is to the EnsureFullCommit step.  This will execute an
index warming action for each batch.  The index refresh will need to
be called for each index defined.

Change-Id: I4de7f8049be8d5ff092c84a42d250ab7c361c492
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Add the ability to create/update, delete and list indexes
to the couchdb layer.

Change-Id: Ia79243a6aff4665cca0f3d144ea24cdd8f312fc4
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

For a while composer has created composite keys in a specific
format and this has worked fine. Since the introduction of a
change to the GetState protocol, GetState has not been returning
information.

Additional investigation showed that the error occurs due to a
bad URL generated when then key contains a colon ":"

Problem is resolved by encoding the path using url EscapedPath()
method instead of the String() method.

Change-Id: I20ad8177301ea845dd234eb867c95b5cde7a908c
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

The problem was that the retrieve of the public
hash from couch state db fails due to a URL
encoding issue of a plus sign in the hashed key.

Keys were already URL encoded using golang URL
encoding, but the encoding skips plus signs in
the path component.  CouchDB unencodes the plus
sign as a space.

The fix is to explicitly URL encode plus character
when interacting with CouchDB.

Also add unit test to verify the fix and test
all other URL special characters.

Also clarified the error message received when
private version does not match public hash.

Change-Id: I9b99d3446542e9eae0196158270205d3ea09db9f
Signed-off-by: David Enyeart <enyeart@us.ibm.com>

CouchDB bulk retrieve and bulk update functions have recently
been added. This change improves readability, maintainability,
serviceability by clarifying code, adding comments, adding
debug.

Change-Id: I552455003ce582288b059f6b434c3d6b93a0688d
Signed-off-by: David Enyeart <enyeart@us.ibm.com>

This is change is part of FAB-2725 CouchDB optimizations

Interactions with CouchDB are currently done individually.
Need to switch to using bulk operations to get optimal performance
from CouchDB. Need to performance test and stress test.

This change implements all changes needed by the commit phase to use
bulk operations in CouchDB.

A subsequent change will implement validation bulk read and caching
of versions and revisions for CouchDB.

Subsequent changes will also be added to retry errors encountered by
the bulk update operation and a configuration setting to limit the
size of the bulk update by number of update documents.

- Add new interface to statedb for bulk operations, the new operations
will only be implemented for databases that can support bulk operations
(only CouchDB)

type BulkOptimizable interface {
	LoadCommittedVersions(keys []*CompositeKey)
	ClearCachedVersions()
}

LoadCommittedVersions will bulk load version and revision numbers into
cache based on the document ID.  Documents missing a version or revision
will be added to the cache with a nil value.

- Add GetVersion method to VersionedDB to enhance batch
validation of batches

GetVersion(namespace string, key string) (*version.Height, error)

- Implement the GetVersion method for CouchDB and goleveldb.  The goleveldb
implementation will be similar to GetState, except it will return the
version.  CouchDB implementation will attempt a cache lookup, then fall
back to a database read if not found.

- Enhance the ApplyUpdates method for CouchDB to incorporate the
ability to retrieve all revision information required for bulk
updates and deletes for CouchDB.

Change-Id: I65f8525a7a13b1ad7b49fda0e357b9a406fba79d
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

This patchset helps clean up the code and docs:

* Fix redundant parenthesese.
* Fix mismatching params in func return.
* Fix variable name collides with imported package name and builtin
  function or reserved words.
* Fix word typos.

Change-Id: I30304c233067ead7e74d18e3caf2604b3ed1490a
Signed-off-by: Baohua Yang <yangbaohua@gmail.com>

A peer panic was encountered connecting to CouchDB on
peer startup.   This occurred in the code after multiple
retries.

The panic was caused by a null pointer when attempting
to read the status code from an http return object.  The
http return object was already checked for an error.

The situation causing the null pointer can only be caused
if the golang http return has a nil for both the response
and the error.   According to golang spec this is not
possible.

An additional check is being made to ensure a good error
message is returned from the CouchDB connection request.

Change-Id: I13740a32231d49faf3a6acf7b37f7bc6e4bba3fa
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

If a state db document update to CouchDB has a http timeout
(currently defaulted at 35s), the transaction may eventually
succeed in CouchDB, but the peer will do a retry using the
previous CouchDB document revision number. The retry will
therefore fail due to revision number conflict. The peer will
think it failed, when in fact it succeeded. Peer panics when
it can't process the block to completion, since subsequent
blocks should not be processed. Need to fix retry logic to
account for this scenario, potentially by introducing a higher
level retry that could get the new revision number for a final
attempt.

Implement a retry in the couchdb SaveDoc/DeleteDoc functions to retry
if the return code is 409 (revision conflict).  If
the return code is 409, then retrieve the current revision
for the document ID and retry.   This logic is separate from
the timeout logic in handleRequest is only used for resolving
document revision conflicts caused by CouchDB timeouts.

Change-Id: I2ef79b63397a8f76e295dc7c562f5cc7f86da73d
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>
Signed-off-by: David Enyeart <enyeart@us.ibm.com>

CouchDB attachments with zero length causes CouchDB to return
successfully but the connection is not available for future
connections.  This is a bug identified in CouchDB.

Zero length attachments are valid.  When an empty byte array
is saved, CouchDB stores the value as a zero length attachment.

The next subsequent client request over the same connection times
out reading headers from server.

This is the CouchDB JIRA item for tracking purposes.
https://issues.apache.org/jira/browse/COUCHDB-3394



The fix in this change checks to see if the attachment in
SaveDoc is zero length.  If the attachment is zero length,
then request close value is set to true.  This setting will
not allow the http connection to be reused and will prevent
connection timeouts.

Change-Id: I1a1a33c44a3b96dfd2bffaf1cec5fe9ece084817
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

CouchDB skip will remain for future use.
Remove the TODO.

Change-Id: I3eb0d4ca70592cdae4bbc4a6a21c78d66990dc82
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

Client and Transport are reusable and thread safe, so
initialize once and reuse across requests. Also close the response
body properly by consuming it fully before closing it to allow
the underlying RoundTripper to reuse the transport for subsequent
requests.

Change-Id: I1d1692ee96db1ed91a0eaccbe81439312c6935ac
Signed-off-by: Balaji Viswanathan <balaji.viswanathan@gmail.com>
Signed-off-by: denyeart <enyeart@us.ibm.com>

This CR updates all loggers throughout the code base to use
`flogging.MustGetLogger`. This function wraps `logging.MustGetLogger`
and tracks the logger modules defined in the system. This enables the
ability to set log levels for modules using regular expressions.
and make it easy to change the levels for any module and all its
submodules with a single command (e.g. gossip, ledger, msp).

Change-Id: If5d3229ea2312adb56fc21bfdafbed3d967cf1df
Signed-off-by: Will Lahti <wtlahti@us.ibm.com>

CouchDB log fills with errors if its system databases are not created.
The system databases should be created upon peer startup,
if they do not yet exist.

This change makes a simple check on creating the state database to
check if the system database exist as well.  If they are missing, then
the system databases are created.

Change-Id: I6dacb8ceefcda8fd0f14a0ef55da921b9160343d
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>

The peer hangs when the CouchDB network connection is unresponsive.
This can be corrected by adding a timeout to the network client
connection used for connecting to CouchDB.  If the connection
times out, then the request to CouchDB will return an error and
the retry logic will take over.

A new entry is added for CouchDB in core.yaml named connectionTimeout

couchDBConfig:
   couchDBAddress: 127.0.0.1:5984
   username:
   password:
   # Number of retries for CouchDB errors
   maxRetries: 3
   # Number of retries for CouchDB errors during peer startup
   maxRetriesOnStartup: 10
   # CouchDB connection timeout (unit: duration, e.g. 60s)
   connectionTimeout: 60s

This is a duration property configurable by the peer.

Change-Id: I5028029e7f303144465c2bfeebf540c5d8e64d82
Signed-off-by: Chris Elder <chris.elder@us.ibm.com>