采购与技术论坛 · 上下游对账的地方

监控归档这个负载，我们一般建议 RAID6 + 热备盘，而不是 RAID10。原因是你的痛点是容量不是随机 IOPS，RAID10 砍掉一半容量在这个行情下成本太高。

重建窗口要压，三个动作：一是重建前先跑一次全盘校验读，把已有坏扇区提前暴露；二是启用重建限速，别为了快把剩余盘压到满载，反而拉高第二块盘故障概率；三是把热备盘做成分组热备（比如每 8 块一个组），而不是全局一个热备 —— 全局热备在大阵列里重建延迟太不可控。

For a surveillance archive workload we would generally recommend RAID 6 with hot spares rather than RAID 10 — your constraint is capacity, not random IOPS, and losing half the capacity costs far too much in this market. To shrink the rebuild window: run a full verification read beforehand so existing bad sectors surface early; enable rebuild rate limiting rather than pinning the surviving drives at full load, which raises second-drive-failure probability; and use grouped hot spares (say one per eight drives) instead of a single global spare, because global spares make rebuild latency far too unpredictable in large arrays.

直接走 DOA，30 天内包来回运费。请把三块的序列号发到 aftersales 邮箱，并附上「换线材换盘位仍不认盘」的说明和一张盘体标签照片 —— 有这三样我们当天就能开单发换新件，不必等你寄回。

另外建议把同批剩余 37 块也抽 2 块做 24 小时烤机。同一批次出现 3 块上电即坏的，通常是运输段的问题（震动或受潮），不是盘本身老化，所以更要确认其余部分没受影响。

Go straight to DOA — return freight both ways is covered within 30 days. Send the three serial numbers to the aftersales mailbox with a note that cables and bays were swapped and a photo of the drive labels; with those three things we can open the case and dispatch replacements the same day without waiting for the return. I would also burn in two of the remaining thirty-seven for 24 hours. Three immediate power-up failures in one batch usually points to transit (shock or moisture) rather than drive aging, so it is worth confirming the rest is unaffected.

按今天的看板，18TB 的每 TB 单价略低，而且这两档我们都有现货、都可以走 DDP 到新加坡。60 块这个量已经进入 50–99 块档，自动 −11.5%。

但更关键的是「同容量替代」：16TB 这一档可选型号更多（希捷 / 西数 / 东芝 / HGST 都有），18TB 目前型号少一些。如果你的项目能接受品牌混用，16TB 的供货确定性反而更高；如果必须单一型号，18TB 要提前锁量。建议把两种方案都发来，我们各出一份报价单对比。

On today's board, 18TB is slightly lower per TB, and both bands are in stock and can ship DDP to Singapore. At sixty units you are already in the 50–99 tier, which is automatic −11.5%. The more important factor is substitutability: the 16TB band has more models available (Seagate, WD, Toshiba and HGST all present), while 18TB currently has fewer. If your project tolerates mixed brands, 16TB gives higher supply certainty; if it must be a single model, lock the 18TB volume early. Send us both options and we will quote each side by side.

我们的口径写在印度尼西亚落地页上：这条禁令是硬门槛，我们不建议在未取得目的地持牌报关行书面意见的情况下发任何一批货进印尼，包括巴淡。巴淡确实存在条件性例外，但「条件」具体怎么落到二手电子设备上，我们看到的公开信息不足以支撑一个可执行的结论。

所以我们的做法是：印尼目前按 FOB 中国香港交货，货到香港为止，进口清关与本地合规责任由买方承担。如果你确实要进印尼，请先拿到报关行的书面意见，我们再按那个意见调整文件与术语。

Our position is written on the Indonesia page: this ban is a hard gate, and we do not recommend shipping into Indonesia — including Batam — without a written opinion from a licensed broker at destination. Conditional exceptions do exist in Batam, but the public information we can see is not sufficient to conclude how those conditions apply to used electronic equipment. So we deliver Indonesia on FOB Hong Kong terms, with the goods handed over in Hong Kong and import clearance and local compliance borne by the buyer. If you do need to enter Indonesia, obtain the broker's written opinion first and we will adapt documents and terms to match.

C7 是接口 CRC 错误，记的是「盘与主机之间传输出错」，不指向盘面。换线还涨，按概率顺序往下查：① 背板接口氧化或针脚变形；② 背板供电不稳（同一背板上多块盘同时涨 C7 基本可以锁定这一条）；③ 阵列卡固件；④ 主板 SATA 控制器。

判断方法很简单：把这 4 块盘换到另一台机器上跑，如果 C7 不再涨，就是原平台的链路问题。如果换机器还在涨，我们再按质保处理。不要急着报修，C7 本身只要没有伴随掉盘或速率异常，通常不影响使用。

C7 counts interface CRC errors — errors in transmission between drive and host — so it does not point at the media. If replacing the cable did not stop it, work down by probability: (1) oxidised or bent pins on the backplane connector; (2) unstable backplane power — multiple drives on the same backplane all climbing at once almost confirms this; (3) RAID card firmware; (4) the motherboard SATA controller. The test is simple: move those four drives to another machine. If C7 stops climbing, the original platform's link is at fault. If it keeps climbing, we handle it under warranty. Do not rush to file an RMA — absent dropouts or abnormal throughput, C7 alone usually does not affect usability.

马来西亚这票我们固定随货四份，加上两份通用文件：① 序列号级功能检验证书；② 装箱单（列明品牌、型号、序列号、制造年份、检验日期）；③ 商业发票（逐项列明型号、数量、单价、申报价值）；④ 外箱标注序列号范围，便于开箱核对。通用两份是原产地文件与运输单据。

关键点是「四份之间的序列号必须完全一致」—— 文件不一致是海关扣货最常见的原因，比文件缺一份更容易出事。发之前我们会把三份文件的序列号做一次交叉核对。

For Malaysia we always ship four destination-specific documents plus two general ones: (1) serial-level functional test certificate; (2) packing list stating brand, model, serial number, year of manufacture and test date; (3) commercial invoice itemising model, quantity, unit price and declared value; and (4) the outer carton marked with the serial range for checking on opening. The two general documents are the origin documentation and transport papers. The critical point is that the serial numbers must match exactly across all four — inconsistent documents are the single most common cause of customs holds, more dangerous than a missing one. Before dispatch we cross-check the serials across all three of the key documents.

氦气流失后盘内阻力上升，同工况下温度会缓慢上升，所以温度确实是氦封盘最实用的观察项之一。但 8 度也可能是机位风道差异 —— 先做一件事：把那块盘和旁边一块互换盘位，跑 2 小时再看。

如果换了位置温度依然高 7–8 度，那基本可以判定不是风道问题，建议隔离观察并记录温度曲线（每小时记一次，连续 48 小时）。持续上升的趋势比绝对值更有意义。如果温度稳定在高位但不再上涨，通常还能正常使用，只是归到 B 级更合适。

As helium escapes, internal drag rises and temperature creeps up under identical conditions, so temperature really is one of the most practical things to watch on helium-sealed drives. But 8°C can also come from airflow position. Do one thing first: swap that drive with a neighbour and run for two hours. If it is still 7–8°C hotter, it is almost certainly not airflow — isolate it and log a temperature curve (hourly for 48 hours). The trend matters more than the absolute value. If it sits stably high without climbing, it is usually still usable, though Grade B is the more honest classification.