redict/tests/unit/pause.tcl

326 lines
9.5 KiB
Tcl
Raw Normal View History

start_server {tags {"pause network"}} {
test "Test read commands are not blocked by client pause" {
r client PAUSE 100000 WRITE
set rd [redis_deferring_client]
$rd GET FOO
$rd PING
$rd INFO
assert_equal [s 0 blocked_clients] 0
r client unpause
$rd close
}
test "Test write commands are paused by RO" {
r client PAUSE 60000 WRITE
set rd [redis_deferring_client]
$rd SET FOO BAR
wait_for_blocked_clients_count 1 50 100
r client unpause
assert_match "OK" [$rd read]
$rd close
}
test "Test special commands are paused by RO" {
r PFADD pause-hll test
r client PAUSE 100000 WRITE
# Test that pfcount, which can replicate, is also blocked
set rd [redis_deferring_client]
$rd PFCOUNT pause-hll
wait_for_blocked_clients_count 1 50 100
# Test that publish, which adds the message to the replication
# stream is blocked.
set rd2 [redis_deferring_client]
$rd2 publish foo bar
wait_for_blocked_clients_count 2 50 100
r client unpause
assert_match "1" [$rd read]
assert_match "0" [$rd2 read]
$rd close
$rd2 close
}
test "Test read/admin mutli-execs are not blocked by pause RO" {
r SET FOO BAR
r client PAUSE 100000 WRITE
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
set rr [redis_client]
assert_equal [$rr MULTI] "OK"
assert_equal [$rr PING] "QUEUED"
assert_equal [$rr GET FOO] "QUEUED"
assert_match "PONG BAR" [$rr EXEC]
assert_equal [s 0 blocked_clients] 0
r client unpause
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
$rr close
}
test "Test write mutli-execs are blocked by pause RO" {
set rd [redis_deferring_client]
$rd MULTI
assert_equal [$rd read] "OK"
$rd SET FOO BAR
assert_equal [$rd read] "QUEUED"
r client PAUSE 60000 WRITE
$rd EXEC
wait_for_blocked_clients_count 1 50 100
r client unpause
assert_match "OK" [$rd read]
$rd close
}
test "Test scripts are blocked by pause RO" {
r client PAUSE 60000 WRITE
set rd [redis_deferring_client]
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
set rd2 [redis_deferring_client]
$rd EVAL "return 1" 0
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
# test a script with a shebang and no flags for coverage
$rd2 EVAL {#!lua
return 1
} 0
wait_for_blocked_clients_count 2 50 100
r client unpause
assert_match "1" [$rd read]
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
assert_match "1" [$rd2 read]
$rd close
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
$rd2 close
}
test "Test RO scripts are not blocked by pause RO" {
r set x y
# create a function for later
r FUNCTION load replace {#!lua name=f1
redis.register_function{
function_name='f1',
callback=function() return "hello" end,
flags={'no-writes'}
}
}
r client PAUSE 6000000 WRITE
set rr [redis_client]
# test an eval that's for sure not in the script cache
assert_equal [$rr EVAL {#!lua flags=no-writes
return 'unique script'
} 0
] "unique script"
# for sanity, repeat that EVAL on a script that's already cached
assert_equal [$rr EVAL {#!lua flags=no-writes
return 'unique script'
} 0
] "unique script"
# test EVAL_RO on a unique script that's for sure not in the cache
assert_equal [$rr EVAL_RO {
return redis.call('GeT', 'x')..' unique script'
} 1 x
] "y unique script"
# test with evalsha
set sha [$rr script load {#!lua flags=no-writes
return 2
}]
assert_equal [$rr EVALSHA $sha 0] 2
# test with function
assert_equal [$rr fcall f1 0] hello
r client unpause
$rr close
}
test "Test read-only scripts in mutli-exec are not blocked by pause RO" {
r SET FOO BAR
r client PAUSE 100000 WRITE
set rr [redis_client]
assert_equal [$rr MULTI] "OK"
assert_equal [$rr EVAL {#!lua flags=no-writes
return 12
} 0
] QUEUED
assert_equal [$rr EVAL {#!lua flags=no-writes
return 13
} 0
] QUEUED
assert_match "12 13" [$rr EXEC]
assert_equal [s 0 blocked_clients] 0
r client unpause
$rr close
}
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
test "Test write scripts in mutli-exec are blocked by pause RO" {
set rd [redis_deferring_client]
set rd2 [redis_deferring_client]
# one with a shebang
$rd MULTI
assert_equal [$rd read] "OK"
$rd EVAL {#!lua
return 12
} 0
assert_equal [$rd read] "QUEUED"
# one without a shebang
$rd2 MULTI
assert_equal [$rd2 read] "OK"
$rd2 EVAL {#!lua
return 13
} 0
assert_equal [$rd2 read] "QUEUED"
r client PAUSE 60000 WRITE
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
$rd EXEC
$rd2 EXEC
wait_for_blocked_clients_count 2 50 100
r client unpause
assert_match "12" [$rd read]
assert_match "13" [$rd2 read]
$rd close
$rd2 close
}
test "Test may-replicate commands are rejected in RO scripts" {
# that's specifically important for CLIENT PAUSE WRITE
assert_error {ERR Write commands are not allowed from read-only scripts. script:*} {
r EVAL_RO "return redis.call('publish','ch','msg')" 0
}
Expose script flags to processCommand for better handling (#10744) The important part is that read-only scripts (not just EVAL_RO and FCALL_RO, but also ones with `no-writes` executed by normal EVAL or FCALL), will now be permitted to run during CLIENT PAUSE WRITE (unlike before where only the _RO commands would be processed). Other than that, some errors like OOM, READONLY, MASTERDOWN are now handled by processCommand, rather than the command itself affects the error string (and even error code in some cases), and command stats. Besides that, now the `may-replicate` commands, PFCOUNT and PUBLISH, will be considered `write` commands in scripts and will be blocked in all read-only scripts just like other write commands. They'll also be blocked in EVAL_RO (i.e. even for scripts without the `no-writes` shebang flag. This commit also hides the `may_replicate` flag from the COMMAND command output. this is a **breaking change**. background about may_replicate: We don't want to expose a no-may-replicate flag or alike to scripts, since we consider the may-replicate thing an internal concern of redis, that we may some day get rid of. In fact, the may-replicate flag was initially introduced to flag EVAL: since we didn't know what it's gonna do ahead of execution, before function-flags existed). PUBLISH and PFCOUNT, both of which because they have side effects which may some day be fixed differently. code changes: The changes in eval.c are mostly code re-ordering: - evalCalcFunctionName is extracted out of evalGenericCommand - evalExtractShebangFlags is extracted luaCreateFunction - evalGetCommandFlags is new code
2022-06-01 07:09:40 -04:00
assert_error {ERR Write commands are not allowed from read-only scripts. script:*} {
r EVAL {#!lua flags=no-writes
return redis.call('publish','ch','msg')
} 0
}
# make sure that publish isn't blocked from a non-RO script
assert_equal [r EVAL "return redis.call('publish','ch','msg')" 0] 0
}
test "Test multiple clients can be queued up and unblocked" {
r client PAUSE 60000 WRITE
set clients [list [redis_deferring_client] [redis_deferring_client] [redis_deferring_client]]
foreach client $clients {
$client SET FOO BAR
}
wait_for_blocked_clients_count 3 50 100
r client unpause
foreach client $clients {
assert_match "OK" [$client read]
$client close
}
}
test "Test clients with syntax errors will get responses immediately" {
r client PAUSE 100000 WRITE
catch {r set FOO} err
assert_match "ERR wrong number of arguments for 'set' command" $err
r client unpause
}
test "Test both active and passive expires are skipped during client pause" {
set expired_keys [s 0 expired_keys]
r multi
r set foo{t} bar{t} PX 10
r set bar{t} foo{t} PX 10
r client PAUSE 50000 WRITE
r exec
wait_for_condition 10 100 {
[r get foo{t}] == {} && [r get bar{t}] == {}
} else {
fail "Keys were never logically expired"
}
# No keys should actually have been expired
assert_match $expired_keys [s 0 expired_keys]
r client unpause
# Force the keys to expire
r get foo{t}
r get bar{t}
# Now that clients have been unpaused, expires should go through
assert_match [expr $expired_keys + 2] [s 0 expired_keys]
}
test "Test that client pause starts at the end of a transaction" {
r MULTI
r SET FOO1{t} BAR
r client PAUSE 60000 WRITE
r SET FOO2{t} BAR
r exec
set rd [redis_deferring_client]
$rd SET FOO3{t} BAR
wait_for_blocked_clients_count 1 50 100
assert_match "BAR" [r GET FOO1{t}]
assert_match "BAR" [r GET FOO2{t}]
assert_match "" [r GET FOO3{t}]
r client unpause
assert_match "OK" [$rd read]
$rd close
}
start_server {tags {needs:repl external:skip}} {
set master [srv -1 client]
set master_host [srv -1 host]
set master_port [srv -1 port]
# Avoid PINGs
$master config set repl-ping-replica-period 3600
r replicaof $master_host $master_port
wait_for_condition 50 100 {
[s master_link_status] eq {up}
} else {
fail "Replication not started."
}
test "Test when replica paused, offset would not grow" {
$master set foo bar
set old_master_offset [status $master master_repl_offset]
wait_for_condition 50 100 {
[s slave_repl_offset] == [status $master master_repl_offset]
} else {
fail "Replication offset not matched."
}
r client pause 100000 write
$master set foo2 bar2
# Make sure replica received data from master
wait_for_condition 50 100 {
[s slave_read_repl_offset] == [status $master master_repl_offset]
} else {
fail "Replication not work."
}
# Replica would not apply the write command
assert {[s slave_repl_offset] == $old_master_offset}
r get foo2
} {}
test "Test replica offset would grow after unpause" {
r client unpause
wait_for_condition 50 100 {
[s slave_repl_offset] == [status $master master_repl_offset]
} else {
fail "Replication not continue."
}
r get foo2
} {bar2}
}
# Make sure we unpause at the end
r client unpause
}