From 4a7357d5e945e5fe89e03b5d65edb45af95b49ee Mon Sep 17 00:00:00 2001 From: Amir Shehata Date: Tue, 3 Jul 2018 17:51:29 -0700 Subject: [PATCH] LU-9120 lnet: set health sensitivity from lnetctl Added an lnetctl command to set the health sensitivity from userspace. lnetctl set health_sensitivity {>0} 0 - turn off health evaluation >0 - sensitivity value not more than 1000 Test-Parameters: forbuildonly Signed-off-by: Amir Shehata Change-Id: Ic9289b06c5c9285a69c1819a33b79e954319a01e Reviewed-on: https://review.whamcloud.com/32779 Reviewed-by: Sonia Sharma Reviewed-by: Olaf Weber Tested-by: Jenkins --- lnet/utils/lnetconfig/liblnetconfig.c | 66 +++++++++++++++++++++++++++++++++-- lnet/utils/lnetconfig/liblnetconfig.h | 24 +++++++++++++ lnet/utils/lnetctl.c | 38 ++++++++++++++++++++ 3 files changed, 126 insertions(+), 2 deletions(-) diff --git a/lnet/utils/lnetconfig/liblnetconfig.c b/lnet/utils/lnetconfig/liblnetconfig.c index 925210e..1527f4a 100644 --- a/lnet/utils/lnetconfig/liblnetconfig.c +++ b/lnet/utils/lnetconfig/liblnetconfig.c @@ -2429,6 +2429,28 @@ int ioctl_set_value(__u32 val, int ioc, char *name, return rc; } +int lustre_lnet_config_hsensitivity(int sen, int seq_no, struct cYAML **err_rc) +{ + int rc = LUSTRE_CFG_RC_NO_ERR; + char err_str[LNET_MAX_STR_LEN]; + char val[LNET_MAX_STR_LEN]; + + snprintf(err_str, sizeof(err_str), "\"success\""); + + snprintf(val, sizeof(val), "%d", sen); + + rc = write_sysfs_file(modparam_path, "lnet_health_sensitivity", val, + 1, strlen(val) + 1); + if (rc) + snprintf(err_str, sizeof(err_str), + "\"cannot configure health sensitivity: %s\"", + strerror(errno)); + + cYAML_build_error(rc, seq_no, ADD_CMD, "health_sensitivity", err_str, err_rc); + + return rc; +} + int lustre_lnet_config_transaction_to(int timeout, int seq_no, struct cYAML **err_rc) { int rc = LUSTRE_CFG_RC_NO_ERR; @@ -3231,6 +3253,31 @@ static int ioctl_show_global_values(int ioc, int seq_no, char *name, data.sv_value, show_rc, err_rc, l_errno); } +int lustre_lnet_show_hsensitivity(int seq_no, struct cYAML **show_rc, + struct cYAML **err_rc) +{ + int rc = LUSTRE_CFG_RC_OUT_OF_MEM; + char val[LNET_MAX_STR_LEN]; + int sen = -1, l_errno = 0; + char err_str[LNET_MAX_STR_LEN]; + + snprintf(err_str, sizeof(err_str), "\"out of memory\""); + + rc = read_sysfs_file(modparam_path, "lnet_health_sensitivity", val, + 1, sizeof(val)); + if (rc) { + l_errno = -errno; + snprintf(err_str, sizeof(err_str), + "\"cannot get health sensitivity: %d\"", rc); + } else { + sen = atoi(val); + } + + return build_global_yaml_entry(err_str, sizeof(err_str), seq_no, + "health_sensitivity", sen, show_rc, + err_rc, l_errno); +} + int lustre_lnet_show_transaction_to(int seq_no, struct cYAML **show_rc, struct cYAML **err_rc) { @@ -4177,7 +4224,8 @@ static int handle_yaml_config_global_settings(struct cYAML *tree, struct cYAML **show_rc, struct cYAML **err_rc) { - struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no; + struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no, + *sen; int rc = 0; seq_no = cYAML_get_object_item(tree, "seq_no"); @@ -4216,6 +4264,13 @@ static int handle_yaml_config_global_settings(struct cYAML *tree, : -1, err_rc); + sen = cYAML_get_object_item(tree, "health_sensitivity"); + if (sen) + rc = lustre_lnet_config_hsensitivity(sen->cy_valueint, + seq_no ? seq_no->cy_valueint + : -1, + err_rc); + return rc; } @@ -4256,7 +4311,8 @@ static int handle_yaml_show_global_settings(struct cYAML *tree, struct cYAML **show_rc, struct cYAML **err_rc) { - struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no; + struct cYAML *max_intf, *numa, *discovery, *retry, *tto, *seq_no, + *sen; int rc = 0; seq_no = cYAML_get_object_item(tree, "seq_no"); @@ -4290,6 +4346,12 @@ static int handle_yaml_show_global_settings(struct cYAML *tree, : -1, show_rc, err_rc); + sen = cYAML_get_object_item(tree, "health_sensitivity"); + if (sen) + rc = lustre_lnet_show_hsensitivity(seq_no ? seq_no->cy_valueint + : -1, + show_rc, err_rc); + return rc; } diff --git a/lnet/utils/lnetconfig/liblnetconfig.h b/lnet/utils/lnetconfig/liblnetconfig.h index 49c71c6..6d67302 100644 --- a/lnet/utils/lnetconfig/liblnetconfig.h +++ b/lnet/utils/lnetconfig/liblnetconfig.h @@ -225,6 +225,30 @@ int lustre_lnet_show_numa_range(int seq_no, struct cYAML **show_rc, struct cYAML **err_rc); /* + * lustre_lnet_config_hsensitivity + * sets the health sensitivity; the value by which to decrement the + * health value of a local or peer NI. If 0 then health is turned off + * + * sen - sensitivity value to configure + * seq_no - sequence number of the request + * err_rc - [OUT] struct cYAML tree describing the error. Freed by + * caller + */ +int lustre_lnet_config_hsensitivity(int sen, int seq_no, struct cYAML **err_rc); + +/* + * lustre_lnet_show_hsensitivity + * show the health sensitivity in the system + * + * seq_no - sequence number of the request + * show_rc - [OUT] struct cYAML tree containing health sensitivity info + * err_rc - [OUT] struct cYAML tree describing the error. Freed by + * caller + */ +int lustre_lnet_show_hsensitivity(int seq_no, struct cYAML **show_rc, + struct cYAML **err_rc); + +/* * lustre_lnet_config_transaction_to * sets the timeout after which a message expires or a timeout event is * propagated for an expired response. diff --git a/lnet/utils/lnetctl.c b/lnet/utils/lnetctl.c index 9a899b9..7633082 100644 --- a/lnet/utils/lnetctl.c +++ b/lnet/utils/lnetctl.c @@ -55,6 +55,7 @@ static int jt_set_large(int argc, char **argv); static int jt_set_numa(int argc, char **argv); static int jt_set_retry_count(int argc, char **argv); static int jt_set_transaction_to(int argc, char **argv); +static int jt_set_hsensitivity(int argc, char **argv); static int jt_add_peer_nid(int argc, char **argv); static int jt_del_peer_nid(int argc, char **argv); static int jt_set_max_intf(int argc, char **argv); @@ -181,6 +182,9 @@ command_t set_cmds[] = { "\t>0 - number of retries\n"}, {"transaction_timeout", jt_set_transaction_to, 0, "Message/Response timeout\n" "\t>0 - timeout in seconds\n"}, + {"health_sensitivity", jt_set_hsensitivity, 0, "sensitivity to failure\n" + "\t0 - turn off health evaluation\n" + "\t>0 - sensitivity value not more than 1000\n"}, { 0, 0, 0, NULL } }; @@ -335,6 +339,34 @@ static int jt_set_numa(int argc, char **argv) return rc; } +static int jt_set_hsensitivity(int argc, char **argv) +{ + long int value; + int rc; + struct cYAML *err_rc = NULL; + + rc = check_cmd(set_cmds, "set", "health_sensitivity", 2, argc, argv); + if (rc) + return rc; + + rc = parse_long(argv[1], &value); + if (rc != 0) { + cYAML_build_error(-1, -1, "parser", "set", + "cannot parse health sensitivity value", &err_rc); + cYAML_print_tree2file(stderr, err_rc); + cYAML_free_tree(err_rc); + return -1; + } + + rc = lustre_lnet_config_hsensitivity(value, -1, &err_rc); + if (rc != LUSTRE_CFG_RC_NO_ERR) + cYAML_print_tree2file(stderr, err_rc); + + cYAML_free_tree(err_rc); + + return rc; +} + static int jt_set_transaction_to(int argc, char **argv) { long int value; @@ -1055,6 +1087,12 @@ static int jt_show_global(int argc, char **argv) goto out; } + rc = lustre_lnet_show_hsensitivity(-1, &show_rc, &err_rc); + if (rc != LUSTRE_CFG_RC_NO_ERR) { + cYAML_print_tree2file(stderr, err_rc); + goto out; + } + if (show_rc) cYAML_print_tree(show_rc); -- 1.8.3.1