This is a regression of commit 4db0017, using stream_socket_recvfrom()
bypasses stream wrappers thus read/write timeouts handled by PHP. The
only way to make them work again is to fetch the raw socket from the
stream resource and use socket_import_stream() which is unfortunately
a PHP 5.4+ function.
This optimization makes it possible to fetch the slots map directly
from the server indicated by the -MOVED response eliminating the need
to use a random node in the pool, which in turn could require Predis
to open a new and useless connection.
This is a more sane default as it allows users to indicate only a few
servers of the whole cluster composition, while it previously required
a more complex configuration of the client using client options.
This feature can be disabled using the "enableClusterNodes()" method.
The command returns the response as is, but it would probably be not a
bad idea to parse it to restructure returned info using named arrays,
making it easier to access single commands and their info.
We will keep it as is for now, and investigate this change later.
CLUSTER SLOTS returns a structured response easier to handle compared
to the bulk response of CLUSTER NODES which must be parsed making the
whole thing more fragile.
CLUSTER SLOTS has been added in Redis 3.0.0b7.
This change is possible because, after a few changes in redis-cluster,
our default cluster strategy used for client side sharding and the one
used for redis-cluster turned out to be exactly the same, except for
the hashing function used to calculate distribution.
Actually some checks used to enforce correctness are redundant in the
context of redis-cluster (e.g. the one used to make sure that keys in
requests performing cross-keys operations will hash to the same slot,
which is performed by the server) so we could also add a more dumb and
permissive cluster strategy that relies on checks performed by Redis.
The only difference between the client-side sharding strategy and the
one used for redis-cluster, aside from the hash function, is in the
rules used for extracting hash tags from keys since empty tags {} are
considered valid by redis-cluster. In v1.0.0 the strategy used for our
client-side sharding cluster will reflect this change, but we will not
change it in v0.8 since it can be considered a breaking changes as it
can potentially affect existing cluster deployments.
Error responses such as -OOM or -ERR on invalid arguments in commands
are returned immediatly instead of +QUEUED when using a transaction,
which is a condition that had not been tested enough. This condition
led to a bug in which Predis was not invalidating the transaction, so
when trying to create a new transaction Redis returned a "-ERR MULTI
calls can not be nested".
This commit fixes#187.
Similarly to the socket-ext based connection using phpiredis, in our
stream based PhpiredisStreamConnection class we should read data from
the stream using stream_socket_recvfrom() instead of fread() because
the latter could block until a timeout is reached when the read buffer
contains less data then the specified length.
IMPORTANT: stream_socket_recvfrom() bypasses stream wrappers which
means that TLS/SSL, as requested by PR #158, won't ever work with
this connection class as the function returns the original encrypted
bytes.
This commit fixes issue #180.
Backported from the master branch.
Many thanks to @rubensayshi for his initial commits on this in #163,
but I had to start from scratch for the master branch first.
Multi-keys operations are not allowed even when keys generate the same
hash but this will probably be supported in later betas of Redis which
means we will basically end up reusing the whole strategy used for
client-side sharding.
Since we cannot rely on the PHP version to detect the availability of
socket_import_stream(), we switched to function_exists(). As an added
bonus, using function_exists() is twice faster.
This method is mostly an utility method which is the main reason for
it bein static, but can be useful to have it available when extending
this class.
This change was inspired by b9b899e (master) even though the original
commit is meant to fix an issue not affecting v0.8.
This oversight didn't end up in a blatant bug only because the correct
handler definition for INFO replaced the old class in the profile.
Backported from 5065541 (master).
Daniele Alessandri
Ian Babrou
Alexey Kupershtokh
thbourlove