We used to copy this value into the server.cluster structure, however this
was not necessary.
The reason why we don't directly use server.cluster->node_timeout is
that things that can be configured via redis.conf need to be directly
available in the server structure as server.cluster is allocated later
only if needed in order to reduce the memory footprint of non-cluster
instances.
In commit d728ec6 it was introduced the concept of sending a ping to
every node not receiving a ping since node_timeout/2 seconds.
However the code was located in a place that was not executed because of
a previous conditional causing the loop to re-iterate.
This caused false positives in nodes availability detection.
The current code is still not perfect as a node may be detected to be in
PFAIL state even if it does not reply for just node_timeout/2 seconds
that is not correct. There is a plan to improve this code ASAP.
When a BGSAVE fails, Redis used to flood itself trying to BGSAVE at
every next cron call, that is either 10 or 100 times per second
depending on configuration and server version.
This commit does not allow a new automatic BGSAVE attempt to be
performed before a few seconds delay (currently 5).
This avoids both the auto-flood problem and filling the disk with
logs at a serious rate.
The five seconds limit, considering a log entry of 200 bytes, will use
less than 4 MB of disk space per day that is reasonable, the sysadmin
should notice before of catastrofic events especially since by default
Redis will stop serving write queries after the first failed BGSAVE.
This fixes issue #849
This commit fixes two corner cases for the TTL command.
1) When the key was already logically expired (expire time older
than current time) the command returned -1 instead of -2.
2) When the key was existing and the expire was found to be exactly 0
(the key was just about to expire), the command reported -1 (that is, no
expire) instead of a TTL of zero (that is, about to expire).
MULTI/EXEC is now propagated to the AOF / Slaves only once we encounter
the first command that is not a read-only one inside the transaction.
The old behavior was to always propagate an empty MULTI/EXEC block when
the transaction was composed just of read only commands, or even
completely empty. This created two problems:
1) It's a bandwidth waste in the replication link and a space waste
inside the AOF file.
2) We used to always increment server.dirty to force the propagation of
the EXEC command, resulting into triggering RDB saves more often
than needed.
Note: even read-only commands may also trigger writes that will be
propagated, when we access a key that is found expired and Redis will
synthesize a DEL operation. However there is no need for this to stay
inside the transaction itself, but only to be ordered.
So for instance something like:
MULTI
GET foo
SET key zap
EXEC
May be propagated into:
DEL foo
MULTI
SET key zap
EXEC
While the DEL is outside the transaction, the commands are delivered in
the right order and it is not possible for other commands to be inserted
between DEL and MULTI.
Redis-tools is a connection of tools no longer mantained that was
intented as a way to economically make sense of Redis in the pre-vmware
sponsorship era. However there was a nice redis-stat utility, this
commit imports one of the functionalities of this tool here in redis-cli
as it seems to be pretty useful.
Usage: redis-cli --stat
The output is similar to vmstat in the format, but with Redis specific
stuff of course.
From the point of view of the monitored instance, only INFO is used in
order to grab data.
This is needed in order to colorize it as next step.
We use conventions in output messages such as
>>> This is an action
*** This is a warning
[ERR] This is an error
[OK] That's fine
And so forth, so that a color will be associated checking the first
three chars.
When a master turns into a slave after a failover event, make sure to
clear the assigned slots before setting up the replication, as a slave
should never claim slots in an explicit way, but just take over the
master slots when replacing its master.
A slave node set this flag for itself when, after receiving authorization
from the majority of nodes, it turns itself into a master.
At the same time now this flag is tested by nodes receiving a PING
message before reconfiguring after a failover event. This makes the
system more robust: even if currently there is no way to manually turn
a slave into a master it is possible that we'll have such a feature in
the future, or that simply because of misconfiguration a node joins the
cluster as master while others believe it's a slave. This alone is now
no longer enough to trigger reconfiguration as other nodes will check
for the PROMOTED flag.
The PROMOTED flag is cleared every time the node is turned back into a
replica of some other node.
Sender flags were not propagated for the sender, but only for nodes in
the gossip section. This is odd and in the next commits we'll need to
get updated flags for the sender node, so this commit adds a new field
in the cluster messages header.
The message header is the same size as we reused some free space that
was marked as 'unused' because of alignment concerns.
So when the failing master node is back in touch with the cluster,
instead of remaining unused it is converted into a replica of the
new master, ready to perform the fail over if the new master node
will fail at some point.
Note that as a side effect clients with stale configuration are now
not an issue as well, as the node converted into a slave will not
accept queries but will redirect clients accordingly.
The code handling a master that turns into a slave or the contrary was
improved in order to avoid repeating the same operations. Also
the readability and conceptual simplicity was improved.
Redis Cluster can cope with a minority of nodes not informed about the
failure of a master in time for some reason (netsplit or node not
functioning properly, blocked, ...) however to wait a few seconds before
to start the failover will make most "normal" failovers simpler as the
FAIL message will propagate before the slave election happens.
server.repl_down_since used to be initialized to the current time at
startup. This is wrong since the replication never started. Clients
testing this filed to check if data is uptodate should never believe
data is recent if we never ever connected to our master.
This fixes cases where the RDB file does exist but can't be accessed for
any reason. For instance, when the Redis process doesn't have enough
permissions on the file.
activeExpireCycle() tries to test just a few DBs per iteration so that
it scales if there are many configured DBs in the Redis instance.
However this commit makes it a bit smarter when one a few of those DBs
are under expiration pressure and there are many many keys to expire.
What we do is to remember if in the last iteration had to return because
we ran out of time. In that case the next iteration we'll test all the
configured DBs so that we are sure we'll test again the DB under
pressure.
Before of this commit after some mass-expire in a given DB the function
tested just a few of the next DBs, possibly empty, a few per iteration,
so it took a long time for the function to reach again the DB under
pressure. This resulted in a lot of memory being used by already expired
keys and never accessed by clients.
This small number of DBs is set to 16 so actually in the default
configuraiton Redis should behave exactly like in the past.
However the difference is that when the user configures a very large
number of DBs we don't do an O(N) operation, consuming a non trivial
amount of CPU per serverCron() iteration.
This is the first step to lower the CPU usage when many databases are
configured. The other is to also process a limited number of DBs per
call in the active expire cycle.
A new server.orig_commands table was added to the server structure, this
contains a copy of the commant table unaffected by rename-command
statements in redis.conf.
A new API lookupCommandOrOriginal() was added that checks both tables,
new first, old later, so that rewriteClientCommandVector() and friends
can lookup commands with their new or original name in order to fix the
client->cmd pointer when the argument vector is renamed.
This fixes the segfault of issue #986, but does not fix a wider range of
problems resulting from renaming commands that actually operate on data
and are registered into the AOF file or propagated to slaves... That is
command renaming should be handled with care.
Usually this does not happens since we trim for " \t\r\n", but if there
are other chars that return true with isspace(), we may end with an
empty argv. Better to handle the condition in an explicit way.
This makes programs not checking the return value for NULL much safer
since with this change:
1) It is still possible to iterate the zero-length result without
crashes.
2) sdssplitargs_free will work against NULL and 0 count.
An empty input string also resulted into the function returning NULL
making it harder for the caller to distinguish between error and empty
string without checking the original input string length.
If we have a master in FAIL state that's reachable again, and apparently
no one is going to serve its slots, clear the FAIL flag and let the
cluster continue with its operations again.
This is the unix time at which we set the FAIL flag for the node.
It is only valid if FAIL is set.
The idea is to use it in order to make the cluster more robust, for
instance in order to revert a FAIL state if it is long-standing but
still slots are assigned to this node, that is, no one is going to fix
these slots apparently.
Usually we try to send just 1 ping every second, however when we detect
we are going to have unreliable failure detection because we can't ping
some node in time, send an additional ping.
This should only happen with very large clusters or when the the node
timeout is set to a very low value.
This should improve things in two ways:
1) Prevent timeouts caused by the execution of long commands.
2) Improve detection of real connection errors.
This is mostly effective only on Linux because of the bogus default
keepalive settings. In Linux we have OS-specific calls to set the
keepalive interval to reasonable values.
As stated in the comment this is usually due to a resharding in progress
so the client should be still redirected to the old node that will
handle the redirection elsewhere.
Before a relatively slow popcount() operation was needed every time we
needed to get the number of slots served by a given cluster node.
Now we just need to check an integer that is taken in sync with the
bitmap.
This cased a segfault in some Linux system and was GCC-specific.
Commit modified by @antirez:
1) Stripped away the part to set the proc title via config for now.
2) Handle initialization of setproctitle only when the replacement
is used.
3) Don't require GCC now that the attribute constructor is no
longer used.
This commit allows Redis to set a process name that includes the binding
address and the port number in order to make operations simpler.
Redis children processes doing AOF rewrites or RDB saving change the
name into redis-aof-rewrite and redis-rdb-bgsave respectively.
This in general makes harder to kill the wrong process because of an
error and makes simpler to identify saving children.
This feature was suggested by Arnaud GRANAL in the Redis Google Group,
Arnaud also pointed me to the setproctitle.c implementation includeed in
this commit.
This feature should work on all the Linux, OSX, and all the three major
BSD systems.
This is not very important as anyway when the function counting the
number of reports is called the cleanup is performed. However with this
change if only part of the nodes that reported the failure will report
the node is back ok, we'll cleanup the older entries ASAP. In complex
split net split scenarios, and when we are dealing with clusters having
nodes in the order of ~ 1000, this can save some CPU.
Not sure why I set a limit to 1 million keys, there is no reason for
this artificial limit, and anyway this is s a stupid limit because it is
already high enough to create latency issues. So let's the users shoot
on their feet because maybe they just actually know what they are doing.
A §Redis Cluster node used to mark a node as failing when itself
detected a failure for that node, and a single acknowledge was received
about the possible failure state.
The new API will be used in order to possible to require that N other
nodes have a PFAIL or FAIL state for a given node for a node to set it
as failing.
This makes us able to avoid allocating the cluster state structure if
cluster is not enabled, but still we can handle the configuration
directive that sets the cluster config filename.
When we are preparing an handshake with the slave we can't touch the
connection buffer as it'll be used to accumulate differences between
the sent RDB file and what arrives next from clients.
So in short we can't use addReply() family functions.
However we just use write(2) because we know that the socket buffer is
empty, since a prerequisite for SYNC to work is that the static buffer
and the output list are empty, and in general it is not expected that a
client SYNCs after doing some heavy I/O with the master.
However a short write connection is explicitly handled to avoid
fragility (we simply close the connection and the slave will retry).
SELECT was still transmitted to slaves using the inline protocol, that
is conceived mostly for humans to type into telnet sessions, and is
notably not understood by redis-cli --slave.
Now the new protocol is used instead.
A Redis master sends PING commands to slaves from time to time: doing
this ensures that even if absence of writes, the master->slave channel
remains active and the slave can feel the master presence, instead of
closing the connection for timeout.
This commit changes the way PINGs are sent to slaves in order to use the
standard interface used to replicate all the other commands, that is,
the function replicationFeedSlaves().
With this change the stream of commands sent to every slave is exactly
the same regardless of their exact state (Transferring RDB for first
synchronization or slave already online). With the previous
implementation the PING was only sent to online slaves, with the result
that the output stream from master to slaves was not identical for all
the slaves: this is a problem if we want to implement partial resyncs in
the future using a global replication stream offset.
TL;DR: this commit should not change the behaviour in practical terms,
but is just something in preparation for partial resynchronization
support.
Before this commit every Redis slave had its own selected database ID
state. This was not actually useful as the emitted stream of commands
is identical for all the slaves.
Now the the currently selected database is a global state that is set to
-1 when a new slave is attached, in order to force the SELECT command to
be re-emitted for all the slaves.
This change is useful in order to implement replication partial
resynchronization in the future, as makes sure that the stream of
commands received by slaves, including SELECT commands, are exactly the
same for every slave connected, at any time.
In this way we could have a global offset that can identify a specific
piece of the master -> slaves stream of commands.
Further details from @antirez:
It was reported by @StopForumSpam on Twitter that the Redis replication
link was strangely using multiple TCP packets for multiple commands.
This wastes a lot of bandwidth and is due to the TCP_NODELAY option we
enable on the socket after accepting a new connection.
However the master -> slave channel is a one-way channel since Redis
replication is asynchronous, so there is no point in trying to reduce
the latency, we should aim to reduce the bandwidth. For this reason this
commit introduces the ability to disable the nagle algorithm on the
socket after a successful SYNC.
This feature is off by default because the delay can be up to 40
milliseconds with normally configured Linux kernels.
In CASE 2, the call sunionDiffGenericCommand will involve the string "srandmember"
> sadd foo one
(integer 1)
> sadd srandmember two
(integer 2)
> srandmember foo 3
1)"one"
2)"two"
When keyspace events are enabled, the overhead is not sever but
noticeable, so this commit introduces the ability to select subclasses
of events in order to avoid to generate events the user is not
interested in.
The events can be selected using redis.conf or CONFIG SET / GET.
decrRefCount used to get its argument as a void* pointer in order to be
used as destructor where a 'void free_object(void*)' prototype is
expected. However this made simpler to introduce bugs by freeing the
wrong pointer. This commit fixes the argument type and introduces a new
wrapper called decrRefCountVoid() that can be used when the void*
argument is needed.
UNSUBSCRIBE and PUNSUBSCRIBE commands are designed to mass-unsubscribe
the client respectively all the channels and patters if called without
arguments.
However when these functions are called without arguments, but there are
no channels or patters we are subscribed to, the old behavior was to
don't reply at all.
This behavior is broken, as every command should always reply.
Also it is possible that we are no longer subscribed to a channels but we
are subscribed to patters or the other way around, and the client should
be notified with the correct number of subscriptions.
Also it is not pretty that sometimes we did not receive a reply at all
in a redis-cli session from these commands, blocking redis-cli trying
to read the reply.
This fixes issue #714.
I don't know how to test for Open Solaris that has support for
backtrace() so for now removing the #ifdef that breaks compilation under
other Solaris flavors.
This commit fixes issue #875 that was caused by the following events:
1) There is an active child doing BGSAVE.
2) flushall is called (or any other condition that makes Redis killing
the saving child process).
3) An error is sensed by Redis as the child exited with an error (killed
by a singal), that stops accepting write commands until a BGSAVE happens
to be executed with success.
Whitelisting SIGUSR1 and making sure Redis always uses this signal in
order to kill its own children fixes the issue.
When a SIGTERM is received Redis schedules a shutdown. However if it
fails to perform the shutdown it must be clear the shutdown_asap flag
otehrwise it will try again and again possibly making the server
unusable.
The Redis Slow Log always used to log the slow commands executed inside
a MULTI/EXEC block. However also EXEC was logged at the end, which is
perfectly useless.
Now EXEC is no longer logged and a test was added to test this behavior.
This fixes issue #759.
Redis pings slaves in "pre-synchronization stage" with newlines. (See
https://github.com/antirez/redis/blob/2.6.9/src/replication.c#L814)
However, redis-cli does not expect this - it sees the newline as the end
of the bulk length line, and ends up returning 0 as bulk the length.
This manifests as the following when running redis-cli:
$ ./src/redis-cli --rdb some_file
SYNC sent to master, writing 0 bytes to 'some_file'
Transfer finished with success.
With this commit, we just ignore leading newlines while reading the bulk
length line.
To reproduce the problem, load enough data into Redis so that the
preparation of the RDB snapshot takes long enough for a ping to occur
while redis-cli is waiting for the data.
Sometimes it is much simpler to debug complex Redis installations if it
is possible to assign clients a name that is displayed in the CLIENT
LIST output.
This is the case, for example, for "leaked" connections. The ability to
provide a name to the client makes it quite trivial to understand what
is the part of the code implementing the client not releasing the
resources appropriately.
Behavior:
CLIENT SETNAME: set a name for the client, or remove the current
name if an empty name is set.
CLIENT GETNAME: get the current name, or a nil.
CLIENT LIST: now displays the client name if any.
Thanks to Mark Gravell for pushing this idea forward.
Issue #828 shows how Redis was not correctly undoing a non-blocking
connection attempt with the previous master when the master was set to a
new address using the SLAVEOF command.
This was also a result of lack of refactoring, so now there is a
function to cancel the non blocking handshake with the master.
The new function is now used when SLAVEOF NO ONE is called or when
SLAVEOF is used to set the master to a different address.
1) The event handler was no restored after a timeout condition if the
command was eventually executed with success.
2) The command was not converted to EVAL in case of errors in the middle
of the execution.
3) Terrible duplication of code without any apparent reason.
REDIS_HZ is the frequency our serverCron() function is called with.
A more frequent call to this function results into less latency when the
server is trying to handle very expansive background operations like
mass expires of a lot of keys at the same time.
Redis 2.4 used to have an HZ of 10. This was good enough with almost
every setup, but the incremental key expiration algorithm was working a
bit better under *extreme* pressure when HZ was set to 100 for Redis
2.6.
However for most users a latency spike of 30 milliseconds when million
of keys are expiring at the same time is acceptable, on the other hand a
default HZ of 100 in Redis 2.6 was causing idle instances to use some
CPU time compared to Redis 2.4. The CPU usage was in the order of 0.3%
for an idle instance, however this is a shame as more energy is consumed
by the server, if not important resources.
This commit introduces HZ as a runtime parameter, that can be queried by
INFO or CONFIG GET, and can be modified with CONFIG SET. At the same
time the default frequency is set back to 10.
In this way we default to a sane value of 10, but allows users to
easily switch to values up to 500 for near real-time applications if
needed and if they are willing to pay this small CPU usage penalty.
Config.h performs endianess detection including OS-specific headers to
define the endianess macros, or when this is not possible, checking the
processor type via ifdefs.
Sometimes when the OS-specific macro is included, only __BYTE_ORDER is
defined, while BYTE_ORDER remains undefined. There is code at the end of
config.h endianess detection in order to define the macros without the
underscore, but it was not working correctly.
This commit fixes endianess detection fixing Redis on Linux / PPC64 and
possibly other systems.
Refactoring performed after issue #801 resolution (see commit
2f87cf8b01) introduced a memory leak that
is fixed by this commit.
I simply forgot to free the new allocated dictionary in the client
structure trusting the output of "make test" on OSX.
However due to changes in the "leaks" utility the test was no longer
testing memory leaks. This problem was also fixed.
Fortunately the CI test running at ci.redis.io spotted the bug in the
valgrind run.
The leak never ended into a stable release.
To store the keys we block for during a blocking pop operation, in the
case the client is blocked for more data to arrive, we used a simple
linear array of redis objects, in the blockingState structure:
robj **keys;
int count;
However in order to fix issue #801 we also use a dictionary in order to
avoid to end in the blocked clients queue for the same key multiple
times with the same client.
The dictionary was only temporary, just to avoid duplicates, but since
we create / destroy it there is no point in doing this duplicated work,
so this commit simply use a dictionary as the main structure to store
the keys we are blocked for. So instead of the previous fields we now
just have:
dict *keys;
This simplifies the code and reduces the work done by the server during
a blocking POP operation.
Sending a command like:
BLPOP foo foo foo foo 0
Resulted into a crash before this commit since the client ended being
inserted in the waiting list for this key multiple times.
This resulted into the function handleClientsBlockedOnLists() to fail
because we have code like that:
if (de) {
list *clients = dictGetVal(de);
int numclients = listLength(clients);
while(numclients--) {
listNode *clientnode = listFirst(clients);
/* server clients here... */
}
}
The code to serve clients used to remove the served client from the
waiting list, so if a client is blocking multiple times, eventually the
call to listFirst() will return NULL or worse will access random memory
since the list may no longer exist as it is removed by the function
unblockClientWaitingData() if there are no more clients waiting for this
list.
To avoid making the rest of the implementation more complex, this commit
modifies blockForKeys() so that a client will be put just a single time
into the waiting list for a given key.
Since it is Saturday, I hope this fixes issue #801.
SDIFF used an algorithm that was O(N) where N is the total number
of elements of all the sets involved in the operation.
The algorithm worked like that:
ALGORITHM 1:
1) For the first set, add all the members to an auxiliary set.
2) For all the other sets, remove all the members of the set from the
auxiliary set.
So it is an O(N) algorithm where N is the total number of elements in
all the sets involved in the diff operation.
Cristobal Viedma suggested to modify the algorithm to the following:
ALGORITHM 2:
1) Iterate all the elements of the first set.
2) For every element, check if the element also exists in all the other
remaining sets.
3) Add the element to the auxiliary set only if it does not exist in any
of the other sets.
The complexity of this algorithm on the worst case is O(N*M) where N is
the size of the first set and M the total number of sets involved in the
operation.
However when there are elements in common, with this algorithm we stop
the computation for a given element as long as we find a duplicated
element into another set.
I (antirez) added an additional step to algorithm 2 to make it faster,
that is to sort the set to subtract from the biggest to the
smallest, so that it is more likely to find a duplicate in a larger sets
that are checked before the smaller ones.
WHAT IS BETTER?
None of course, for instance if the first set is much larger than the
other sets the second algorithm does a lot more work compared to the
first algorithm.
Similarly if the first set is much smaller than the other sets, the
original algorithm will less work.
So this commit makes Redis able to guess the number of operations
required by each algorithm, and select the best at runtime according
to the input received.
However, since the second algorithm has better constant times and can do
less work if there are duplicated elements, an advantage is given to the
second algorithm.
The idea is to be able to identify a build in a unique way, so for
instance after a bug report we can recognize that the build is the one
of a popular Linux distribution and perform the debugging in the same
environment.
1) We no longer test location by location, otherwise the CPU write cache
completely makes our business useless.
2) We still need a memory test that operates in steps from the first to
the last location in order to never hit the cache, but that is still
able to retain the memory content.
This was tested using a Linux box containing a bad memory module with a
zingle bit error (always zero).
So the final solution does has an error propagation step that is:
1) Invert bits at every location.
2) Swap adiacent locations.
3) Swap adiacent locations again.
4) Invert bits at every location.
5) Swap adiacent locations.
6) Swap adiacent locations again.
Before and after these steps, and after step 4, a CRC64 checksum is computed.
If the three CRC64 checksums don't match, a memory error was detected.
EVALSHA used to crash if the SHA1 was not lowercase (Issue #783).
Fixed using a case insensitive dictionary type for the sha -> script
map used for replication of scripts.
After the transcation starts with a MULIT, the previous behavior was to
return an error on problems such as maxmemory limit reached. But still
to execute the transaction with the subset of queued commands on EXEC.
While it is true that the client was able to check for errors
distinguish QUEUED by an error reply, MULTI/EXEC in most client
implementations uses pipelining for speed, so all the commands and EXEC
are sent without caring about replies.
With this change:
1) EXEC fails if at least one command was not queued because of an
error. The EXECABORT error is used.
2) A generic error is always reported on EXEC.
3) The client DISCARDs the MULTI state after a failed EXEC, otherwise
pipelining multiple transactions would be basically impossible:
After a failed EXEC the next transaction would be simply queued as
the tail of the previous transaction.
We use this new bio.c feature in order to stop our I/O threads if there
is a memory test to do on crash. In this case we don't want anything
else than the main thread to run, otherwise the other threads may mess
with the heap and the memory test will report a false positive.
